This article was published on another Kaggle forum, but I decided to share it with you here as well. These are my lessons learned from various Kaggle competitions. My point of view. Maybe some of you will find it interesting, useful? Who knows?
\nI remember the day I joined Kaggle … I didn't know what I was doing …. I was looking for help, for explanations … Everything seemed so complicated to me. And when I looked at hundreds of lines of Kagglers code … I couldn't believe that I would write such things one day.
\nI appereciate all of you who help me grow here. Thank you!
", "messages": [ { "id": 1632877, "postDate": "2021-12-30T07:28:43.380Z", "content": "This article was published on another Kaggle forum, but I decided to share it with you here as well. These are my lessons learned from various Kaggle competitions. My point of view. Maybe some of you will find it interesting, useful? Who knows?
\nI remember the day I joined Kaggle … I didn't know what I was doing …. I was looking for help, for explanations … Everything seemed so complicated to me. And when I looked at hundreds of lines of Kagglers code … I couldn't believe that I would write such things one day.
\nI appereciate all of you who help me grow here. Thank you!
", "rawMarkdown": "This article was published on another Kaggle forum, but I decided to share it with you here as well. These are my lessons learned from various Kaggle competitions. My point of view. Maybe some of you will find it interesting, useful? Who knows?\n\nI remember the day I joined Kaggle ... I didn't know what I was doing .... I was looking for help, for explanations ... Everything seemed so complicated to me. And when I looked at hundreds of lines of Kagglers code ... I couldn't believe that I would write such things one day.\n\n- Practice, practice, practice. Practicing make you perfect. Minimize talking about AI ... maximize practice. My experience grows with the number of lines of code written (it's thousands, thousands ... lines and hundreds of notebooks - I remember one competition where I generated over 60 different notebooks).\n- Data, data ... data ... 90% ... 9% model, 1% tuning - before you move on you must know everything about dataset. It is a broad topic ... it has many elements ... but it is the basis.\n- The solutions are usually simpler than we think.\n- I like to be surprised by clever solutions proposed by smarter than myself ... it's such a nice feeling when someone breaks your head ... oh, there are many of them on Kaggle ... respect!\n- I love the moments of Kaggle breakthrough ... when the whole world stands still for a few days, no one improves the result, then someone shares an insight, a solution and the whole world moves forward ... it shows what power lies in cooperation and sharing. It is such behavior that builds progress and innovation. Breakthroughs like this are often experienced at Kaggle competitions ... valuable experience.\n- Usually the best answer is \"I don't know\" ... we have to check it and conduct many experiments to prove the thesis. If someone, without conducting an experiment, says that \"it's like this\" then 98% is wrong.\n- I like the moment when I make a submission on Kaggle and ... wait for the results. I learned not to build expectations ... so often I greeted the \"goose\" to see later that the solution is much worse than I thought ....\n- Stay up to date with SoTA ... and ML / AI news ...\n- Learn the basics - master Python ... numpy, pandas. Then Keras / Tensorflow or / and Pytorch ... classification (implement some basic architectures from scratch), then add a head to it and make a simple detector .... and create, experiment and ... have fun!\n- Talk to people ... with enthusiasts ... on Kaggle I met people from all over the world ... France, Germany, Japan, China, Korea, USA, Turkey .... Poland ... we are still trying to figure out how to \"push\" the limits \", how to jump over another 0.001 in the result ... Sometimes it is just one hyperparameter... other times hundreds lines of code ...\n- look for a mentor .... someone from whom you will be able to learn how to make a good solution even better .... \n- if you want to develop, you MUST share ... your knowledge and solutions ...\n- is it possible to be in TOP100 kagglers in one year ... yes it is ... :) \n- I often see that people optimize model at the beggining .... Why? How to optimize something which is far from good model? How to optimize unstable system. As a result, they get worse results than if they ran the model on the default parameters. This is just last part of the game. Fight for the best model without tuning. Do this one last.\n- I often see EDA .... without \"so what\" .... just ... good looking charts ... absolutely oudstanding charts .... but ... there is no conclusion ... \"so what\" ... \n- We are often tempted to use what we already know. Since I feel good in neural networks, I start with them again. However, this is a comfort zone. Do something completely different. Learn a new thing. Expand your set of available tools.\n- Why is my notebook rated so low? Why does it only have 20 votes? Check the number of forks. You will discover its true value and usefulness. Sometimes my notebook only has 20-30 votes or it lacks a few points to gold ... I look at the number of forks and I can't believe ...\n- Stay focused .... I know that daily there is a lot new ideas in competition (or your project) ... stay focused ... “If you chase two rabbits...You will not catch either one.” Luckily Kaggle provide only 5 submission daily ... this limitation is for thinking ... not Kaggle's malice.\n\nI appereciate all of you who help me grow here. Thank you!", "votes": 236 }, { "id": 1637011, "postDate": "2022-01-03T13:23:26.220Z", "content": "i can add one more point:
\nfor example, one may note that there are more FPs at the bottom of the images. These are FP that are similar partial starfish that are out of the image frame. But there is a trick. Because we are dealing with video, the starfish will move from the center of the image to the bottom edge of the image. i.e. there don't suddenly appears at bottom of images.
\nso if you have low confidence detection of \"partial starfish\" at the bottom of image, they are more likely to be FP if you cannot track from center of the image from previous frames.
\n(this is a heuristics method. if you want ML solution, you can add temporal and coordinate features to your detection/tracking model)
\nmany or tricks can be easily discovered by observations
", "rawMarkdown": "i can add one more point:\n\n- always print out / draw your prediction results and errors and observe\n\nfor example, one may note that there are more FPs at the bottom of the images. These are FP that are similar partial starfish that are out of the image frame. But there is a trick. Because we are dealing with video, the starfish will move from the center of the image to the bottom edge of the image. i.e. there don't suddenly appears at bottom of images.\n\nso if you have low confidence detection of \"partial starfish\" at the bottom of image, they are more likely to be FP if you cannot track from center of the image from previous frames.\n\n(this is a heuristics method. if you want ML solution, you can add temporal and coordinate features to your detection/tracking model)\n\nmany or tricks can be easily discovered by observations", "votes": 17 }, { "id": 1633960, "postDate": "2021-12-31T08:12:45.377Z", "content": "Thank you for this advice! So many sites claim to be a means to community, but don't invite it in any ways beyond opening a forum. Your post really helps me understand how central humbly learning from other Kagglers is at every step. Thank you!
", "rawMarkdown": "Thank you for this advice! So many sites claim to be a means to community, but don't invite it in any ways beyond opening a forum. Your post really helps me understand how central humbly learning from other Kagglers is at every step. Thank you!", "votes": 3, "replies": [ { "id": 1634131, "postDate": "2021-12-31T11:17:51.590Z", "content": "Thank you @shonmeckfessel for your feedback. I really appreciate that. I'm not the best. I am not even very good, but … I try to share my works and improve them to learn new things overnight. And I like to have fun. That's why I usually treat the competition as entertainment, not stress and a race for $$$. For me, the number one goal is to have fun and develop myself.
", "rawMarkdown": "Thank you @shonmeckfessel for your feedback. I really appreciate that. I'm not the best. I am not even very good, but ... I try to share my works and improve them to learn new things overnight. And I like to have fun. That's why I usually treat the competition as entertainment, not stress and a race for $$$. For me, the number one goal is to have fun and develop myself.", "votes": 2 } ] }, { "id": 1646361, "postDate": "2022-01-11T16:54:47.187Z", "content": "This is very useful.
", "rawMarkdown": "This is very useful.", "votes": 1, "replies": [ { "id": 1650184, "postDate": "2022-01-14T21:48:28.467Z", "content": "Thank you! Have a nic experience with Kaggle and supporting people! This is improtant here. I have to write it as a next lessons learned. Competing is important but … having people supporting is key.
", "rawMarkdown": "Thank you! Have a nic experience with Kaggle and supporting people! This is improtant here. I have to write it as a next lessons learned. Competing is important but ... having people supporting is key." } ] }, { "id": 1645686, "postDate": "2022-01-11T06:22:46.673Z", "content": "thanks for sharing. That's very useful for me. The sharing is very inspiring.
", "rawMarkdown": "thanks for sharing. That's very useful for me. The sharing is very inspiring.", "votes": 1 }, { "id": 1645192, "postDate": "2022-01-10T18:42:46.607Z", "content": "Quite useful!
", "rawMarkdown": "Quite useful!", "votes": 1 }, { "id": 1644941, "postDate": "2022-01-10T16:09:31.937Z", "content": "Thanks for sharing your experience.Hope I can profit from it!!!👍
", "rawMarkdown": "Thanks for sharing your experience.Hope I can profit from it!!!👍", "votes": 1 }, { "id": 1644695, "postDate": "2022-01-10T12:18:30.753Z", "content": "Thanks! Many great tips for any beginner in machine learning.
", "rawMarkdown": "Thanks! Many great tips for any beginner in machine learning.", "votes": 1 }, { "id": 1644572, "postDate": "2022-01-10T10:55:26.820Z", "content": "You can add another:
\nOne tip that wasn't added:-
\nWhile keeping to the above listed suggestions, it should be uppermost in our minds that participation in competitions is just to gain valuable experience.
\nThe objective is not winning or getting a medal.
Inciteful @remekkinas
\nThank you for sharing!
Good points
", "rawMarkdown": "Good points ", "votes": 1 }, { "id": 1643333, "postDate": "2022-01-09T10:20:50.700Z", "content": "your words are going to build our career
", "rawMarkdown": "your words are going to build our career ", "votes": 1 }, { "id": 1643233, "postDate": "2022-01-09T07:59:18.583Z", "content": "Thanks for sharing. This is very useful info.
\nI'm a kaggle beginner, so I can't read the source code of object detection (e.g. YOLOX) perfectly.
\nBut, I hope I can read the source code of various object detection methods someday.
Thank you so much for sharing. I'm a beginner in ML. Would you like to give me some advice?
", "rawMarkdown": "Thank you so much for sharing. I'm a beginner in ML. Would you like to give me some advice?", "votes": 1 }, { "id": 1642983, "postDate": "2022-01-08T21:14:08.753Z", "content": "Thanks for sharing! As a newbie this post is very helpful. I will commit to practice even more, maybe one day I can be be as confident and reach your pace.
", "rawMarkdown": "Thanks for sharing! As a newbie this post is very helpful. I will commit to practice even more, maybe one day I can be be as confident and reach your pace. ", "votes": 1 }, { "id": 1642828, "postDate": "2022-01-08T18:04:31.607Z", "content": "I can relate to one of the points, solution is simple than you think. I tend to take unnecessary stress trying to write a solution and making it complicated.
", "rawMarkdown": "I can relate to one of the points, solution is simple than you think. I tend to take unnecessary stress trying to write a solution and making it complicated. ", "votes": 1 }, { "id": 1642610, "postDate": "2022-01-08T14:29:53.857Z", "content": "@remekkinas Thank you so much for sharing your 2021. It is a great list! I do have one question. Can you share where you \"Stay up to date with SoTA … and ML / AI news …\" outside of kaggle. Thank you.
", "rawMarkdown": "@remekkinas Thank you so much for sharing your 2021. It is a great list! I do have one question. Can you share where you \"Stay up to date with SoTA … and ML / AI news …\" outside of kaggle. Thank you.", "votes": 1 }, { "id": 1642423, "postDate": "2022-01-08T10:55:20.587Z", "content": "Great post. Hope I can do a transfer learning from this ? :)
", "rawMarkdown": "Great post. Hope I can do a transfer learning from this ? :)", "votes": 1 }, { "id": 1642328, "postDate": "2022-01-08T09:17:43.233Z", "content": "Thanks for sharing your experience，I hope I can catch up with your pace ！
", "rawMarkdown": "Thanks for sharing your experience，I hope I can catch up with your pace ！", "votes": 1, "replies": [ { "id": 1642329, "postDate": "2022-01-08T09:20:38.237Z", "content": "I am sure you are able to do this …. just spend as much time as possible on Kaggle.
", "rawMarkdown": "I am sure you are able to do this .... just spend as much time as possible on Kaggle." } ] }, { "id": 1642285, "postDate": "2022-01-08T08:41:02.333Z", "content": "This one is awesome !!!
", "rawMarkdown": "This one is awesome !!!", "votes": 1 }, { "id": 1641970, "postDate": "2022-01-07T21:53:38.647Z", "content": "Thank you for sharing! As a newbie, this is incredibly helpful!
", "rawMarkdown": "Thank you for sharing! As a newbie, this is incredibly helpful!", "votes": 1 }, { "id": 1641804, "postDate": "2022-01-07T18:07:26.320Z", "content": "Practice, consistency, and hard work are three essential things for improving. Keep growing @remekkinas
", "rawMarkdown": "Practice, consistency, and hard work are three essential things for improving. Keep growing @remekkinas ", "votes": 1, "replies": [ { "id": 1642330, "postDate": "2022-01-08T09:20:51.963Z", "content": "Thank you very much!
", "rawMarkdown": "Thank you very much!" } ] }, { "id": 1641335, "postDate": "2022-01-07T10:43:12.390Z", "content": "Respect!!!
", "rawMarkdown": "Respect!!!", "votes": 1 }, { "id": 1640642, "postDate": "2022-01-06T16:59:53.917Z", "content": "Thank You for the advice! In the post you talked about looking for a mentor, can you please describe, how should i look for a mentor?
", "rawMarkdown": "Thank You for the advice! In the post you talked about looking for a mentor, can you please describe, how should i look for a mentor?", "votes": 1, "replies": [ { "id": 1642332, "postDate": "2022-01-08T09:22:50.753Z", "content": "To be honest … the best mentor is …… Kaggle …. I was trying to find somebody but unfortunately this is really hard (expensive or people do not want to help …. or … I have not found such person - read I am not good at looging for people).
", "rawMarkdown": "To be honest ... the best mentor is ...... Kaggle .... I was trying to find somebody but unfortunately this is really hard (expensive or people do not want to help .... or ... I have not found such person - read I am not good at looging for people)." } ] }, { "id": 1640420, "postDate": "2022-01-06T13:20:53.747Z", "content": "So inspiring! Awesome 👍
", "rawMarkdown": "So inspiring! Awesome 👍", "votes": 1 }, { "id": 1640301, "postDate": "2022-01-06T11:07:50.130Z", "content": "This is so true. I started out with reading a lot of books and online courses, but when faced with a real problem I was stuck. No clue on how to proceed and where to even begin. Now I need to face my fears and accept that only practice makes perfect, if perfect is ever achievable, and just dig in and write code every day, learn from my mistakes and do better the next day. So thank you @remekkinas for the post and inspiration.
", "rawMarkdown": "This is so true. I started out with reading a lot of books and online courses, but when faced with a real problem I was stuck. No clue on how to proceed and where to even begin. Now I need to face my fears and accept that only practice makes perfect, *if perfect is ever achievable*, and just dig in and write code every day, learn from my mistakes and do better the next day. So thank you @remekkinas for the post and inspiration.", "votes": 1, "replies": [ { "id": 1642334, "postDate": "2022-01-08T09:25:06.727Z", "content": "@tadejsitar thank you for your feedback! I have the same opinion … reading book is knowledge … but knowledge is not enough. To solve problems, we must have skills and practical experience.
", "rawMarkdown": "@tadejsitar thank you for your feedback! I have the same opinion ... reading book is knowledge ... but knowledge is not enough. To solve problems, we must have skills and practical experience.", "votes": 1 } ] }, { "id": 1640175, "postDate": "2022-01-06T08:55:12.650Z", "content": "Thanks for Sharing, It's very informative.
", "rawMarkdown": "Thanks for Sharing, It's very informative.", "votes": 1 }, { "id": 1640028, "postDate": "2022-01-06T06:04:26.630Z", "content": "i respect you!!
", "rawMarkdown": "i respect you!!", "votes": 1 }, { "id": 1639973, "postDate": "2022-01-06T05:10:39.480Z", "content": "Inspiring post @remekkinas!!!
\nYes, we can grow in Kaggle by practice, practice, and more practice… share our knowledge and learn from clever solutions!!!
", "rawMarkdown": "Inspiring post @remekkinas!!! \n\nYes, we can grow in Kaggle by practice, practice, and more practice... share our knowledge and learn from clever solutions!!!", "votes": 1, "replies": [ { "id": 1640045, "postDate": "2022-01-06T06:16:56.503Z", "content": "Yes! Thank you for your comment!
", "rawMarkdown": "Yes! Thank you for your comment!" } ] }, { "id": 1639933, "postDate": "2022-01-06T04:24:15.447Z", "content": "thank you for sharing your thought. it make me on fire haha
", "rawMarkdown": "thank you for sharing your thought. it make me on fire haha", "votes": 1 }, { "id": 1639870, "postDate": "2022-01-06T02:23:36.677Z", "content": "This one is aweome !!!
", "rawMarkdown": "This one is aweome !!!", "votes": 1 }, { "id": 1639838, "postDate": "2022-01-06T00:39:35.733Z", "content": "Thank you for the insightful post! As someone who is just starting out this is great info. My hope is to make it to Master by the end of the year. Hopefully, I can look back on this comment and be happy with my progress. And that will be in part thanks to your insight!
", "rawMarkdown": "Thank you for the insightful post! As someone who is just starting out this is great info. My hope is to make it to Master by the end of the year. Hopefully, I can look back on this comment and be happy with my progress. And that will be in part thanks to your insight!", "votes": 1, "replies": [ { "id": 1640393, "postDate": "2022-01-06T12:54:49.353Z", "content": "I think you are able to to this. But sharing, sharing, sharing. Spend time to learn Kaggle and ML.
", "rawMarkdown": "I think you are able to to this. But sharing, sharing, sharing. Spend time to learn Kaggle and ML. " } ] }, { "id": 1639091, "postDate": "2022-01-05T11:15:15.670Z", "content": "Hope to be \"niubi \" as you after 1 year , now i am just a little little rookie .
", "rawMarkdown": "Hope to be \"niubi \" as you after 1 year , now i am just a little little rookie .", "votes": 1 }, { "id": 1638746, "postDate": "2022-01-05T01:42:44.820Z", "content": "Respect!!!!!
", "rawMarkdown": "Respect!!!!!", "votes": 1 }, { "id": 1638551, "postDate": "2022-01-04T20:21:10.480Z", "content": "Thnaks for sharing
", "rawMarkdown": "Thnaks for sharing", "votes": 1 }, { "id": 1638270, "postDate": "2022-01-04T15:23:53.150Z", "content": "There are many points which I know and many which I don't know. Thank you for sharing it
", "rawMarkdown": "There are many points which I know and many which I don't know. Thank you for sharing it ", "votes": 1 }, { "id": 1638017, "postDate": "2022-01-04T11:27:35.857Z", "content": "This is what a New Year motivation i needed!! i will try to get close enough to these goals!! THANKS A TON👍
", "rawMarkdown": "This is what a New Year motivation i needed!! i will try to get close enough to these goals!! THANKS A TON👍", "votes": 1 }, { "id": 1637717, "postDate": "2022-01-04T06:27:09.977Z", "content": "Respect!!!
", "rawMarkdown": "Respect!!!", "votes": 1 }, { "id": 1637205, "postDate": "2022-01-03T16:48:26.140Z", "content": "Amazing 🔥🔥
", "rawMarkdown": "Amazing 🔥🔥", "votes": 1 }, { "id": 1637156, "postDate": "2022-01-03T16:06:56.043Z", "content": "Thanks for the insights, great motivation!
", "rawMarkdown": "Thanks for the insights, great motivation!", "votes": 1 }, { "id": 1637025, "postDate": "2022-01-03T13:41:11.037Z", "content": "Really inspiring
", "rawMarkdown": "Really inspiring", "votes": 1, "replies": [ { "id": 1637052, "postDate": "2022-01-03T14:01:55.493Z", "content": "Thank you! 🙏
", "rawMarkdown": "Thank you! 🙏" } ] }, { "id": 1636459, "postDate": "2022-01-02T22:59:36.430Z", "content": "This post gives me the motivation to push forward my Kaggle journey. Thank you @remekkinas
", "rawMarkdown": "This post gives me the motivation to push forward my Kaggle journey. Thank you @remekkinas ", "votes": 1, "replies": [ { "id": 1636654, "postDate": "2022-01-03T06:33:22.417Z", "content": "You are welcome :)
", "rawMarkdown": "You are welcome :)", "votes": 1 } ] }, { "id": 1635960, "postDate": "2022-01-02T12:54:50.050Z", "content": "Thank you for sharing this insightful article! It is really helpful.
", "rawMarkdown": "Thank you for sharing this insightful article! It is really helpful.", "votes": 1, "replies": [ { "id": 1635970, "postDate": "2022-01-02T13:03:43.597Z", "content": "Thnak you very much for you feedback! 👍😃
", "rawMarkdown": "Thnak you very much for you feedback! 👍😃", "votes": 1 } ] }, { "id": 1635945, "postDate": "2022-01-02T12:21:28.740Z", "content": "Thank you for your very helpful insight
", "rawMarkdown": "Thank you for your very helpful insight", "votes": 1, "replies": [ { "id": 1635946, "postDate": "2022-01-02T12:24:47.563Z", "content": "You are welcome! :)
", "rawMarkdown": "You are welcome! :)", "votes": 1 } ] }, { "id": 1635833, "postDate": "2022-01-02T08:46:47.067Z", "content": "Hello, thank you for this article with minor \"best practices\". I only beginner in kaggle and this article help me understand what i need for get better result :)
", "rawMarkdown": "Hello, thank you for this article with minor \"best practices\". I only beginner in kaggle and this article help me understand what i need for get better result :)", "votes": 1, "replies": [ { "id": 1636655, "postDate": "2022-01-03T06:33:51.643Z", "content": "This is my experience but … each day collecting new one :)
", "rawMarkdown": "This is my experience but ... each day collecting new one :)", "votes": 1 } ] }, { "id": 1635796, "postDate": "2022-01-02T07:48:35.833Z", "content": "Thank you for this insightful advice. Any tips // specific resources for staying up-to-date on ML // AI news?
", "rawMarkdown": "Thank you for this insightful advice. Any tips // specific resources for staying up-to-date on ML // AI news?\n", "votes": 1, "replies": [ { "id": 1635800, "postDate": "2022-01-02T07:52:43.597Z", "content": "I use twitter to learn about new ai/ml solutions. Not joking.
", "rawMarkdown": "I use twitter to learn about new ai/ml solutions. Not joking." } ] }, { "id": 1635741, "postDate": "2022-01-02T06:34:32.137Z", "content": "Great article. Very valuable for newbies.
", "rawMarkdown": "Great article. Very valuable for newbies.", "votes": 1, "replies": [ { "id": 1635805, "postDate": "2022-01-02T07:57:26.263Z", "content": "Thank you!
", "rawMarkdown": "Thank you!" } ] }, { "id": 1634836, "postDate": "2022-01-01T08:36:14.930Z", "content": "It is precisely the kind of post I wanted to start my new year reading. Thanks for sharing. It is pretty informative!
", "rawMarkdown": "It is precisely the kind of post I wanted to start my new year reading. Thanks for sharing. It is pretty informative!", "votes": 1, "replies": [ { "id": 1634849, "postDate": "2022-01-01T08:58:29.010Z", "content": "Thank you very much for such kind feedback :) Have a outstanding 2022 :)
", "rawMarkdown": "Thank you very much for such kind feedback :) Have a outstanding 2022 :)", "votes": 1 } ] }, { "id": 1633670, "postDate": "2021-12-31T01:15:08.837Z", "content": "Totally agreed. I have just joined recently as well. but have learnt a lot since then. Thanks for all people who willing to teach and share idea here.
", "rawMarkdown": "Totally agreed. I have just joined recently as well. but have learnt a lot since then. Thanks for all people who willing to teach and share idea here. ", "votes": 1, "replies": [ { "id": 1634118, "postDate": "2021-12-31T11:11:19.313Z", "content": "Thank you @hawkeat! I personally agree with you. Sharing is essence of portal like this one.
", "rawMarkdown": "Thank you @hawkeat! I personally agree with you. Sharing is essence of portal like this one. " } ] }, { "id": 1633560, "postDate": "2021-12-30T22:44:02.193Z", "content": "Thank you so much for sharing your know-hows, ideas and codes at this competition.
\nI've started Kaggle just a week ago and never touched YOLO,
\nbut now I can train my YOLOX model thanks to your posts!
Have a good new year!
", "rawMarkdown": "Thank you so much for sharing your know-hows, ideas and codes at this competition.\n\nI've started Kaggle just a week ago and never touched YOLO, \nbut now I can train my YOLOX model thanks to your posts!\n\nHave a good new year!", "votes": 1, "replies": [ { "id": 1634125, "postDate": "2021-12-31T11:14:19.460Z", "content": "@roniheka ❤️❤️❤️ Thank you!!! I really like stories like this one ….
\n\"yesterday I did not know how to start with object detection\"
\n\"today I learned YOLO and created first object detection model\"
\n\"tomorrow I will develop my object detection solution, find way how to improve my solution\"
I can see progress. All is about progress and continuous improvement.
", "rawMarkdown": "@roniheka ❤️❤️❤️ Thank you!!! I really like stories like this one .... \n\n\"yesterday I did not know how to start with object detection\"\n\"today I learned YOLO and created first object detection model\"\n\"tomorrow I will develop my object detection solution, find way how to improve my solution\"\n\nI can see progress. All is about progress and continuous improvement. ", "votes": 3 } ] }, { "id": 1643241, "postDate": "2022-01-09T08:20:08.683Z", "content": "Thanks for these significant key points!
", "rawMarkdown": "Thanks for these significant key points!", "votes": 2 }, { "id": 1639192, "postDate": "2022-01-05T13:02:43.557Z", "content": "As a newcomer to this world of data analytics, reading your post was helpful. I really like the collaboration mindset which is seems prevalent here.
", "rawMarkdown": "As a newcomer to this world of data analytics, reading your post was helpful. I really like the collaboration mindset which is seems prevalent here.", "votes": 2 }, { "id": 1638399, "postDate": "2022-01-04T17:31:25.323Z", "content": "Very valuable for newbies. !Thank you…
", "rawMarkdown": "Very valuable for newbies. !Thank you...", "votes": 2 }, { "id": 2309898, "postDate": "2023-06-20T01:00:14.990Z", "content": "Practice, practice and practice.
", "rawMarkdown": "Practice, practice and practice." }, { "id": 1748828, "postDate": "2022-04-08T03:28:50.417Z", "content": "I'm gonna save and bookmark these thoughtful words and check it from time to time as long as I'm on Kaggle ❤🔥❤🔥❤
", "rawMarkdown": "I'm gonna save and bookmark these thoughtful words and check it from time to time as long as I'm on Kaggle ❤🔥❤🔥❤" }, { "id": 1722534, "postDate": "2022-03-14T16:18:37.247Z", "content": "Practice, practice, practice.
\nStay focused
\nMany thanks for your great tips
Thank you so much for the tips! I'm curious, what was your background when starting on Kaggle? How much coding experience did you have?
", "rawMarkdown": "Thank you so much for the tips! I'm curious, what was your background when starting on Kaggle? How much coding experience did you have?", "replies": [ { "id": 1704859, "postDate": "2022-02-25T22:54:13.693Z", "content": "I am Computer Science by education (AI). Coding … a lot but mainly in C/C++ Then I learned R …. and decided to switch to Python which is my favourite programming language for AI/ML and prototyping. Still use C/C++.
", "rawMarkdown": "I am Computer Science by education (AI). Coding ... a lot but mainly in C/C++ Then I learned R .... and decided to switch to Python which is my favourite programming language for AI/ML and prototyping. Still use C/C++.", "votes": 1 } ] }, { "id": 1683905, "postDate": "2022-02-10T06:14:32.017Z", "content": "Great insights! Thanks for sharing. Do you have any suggestions for staying on top of latest ML/AI news? Any journals you would recommend?
", "rawMarkdown": "Great insights! Thanks for sharing. Do you have any suggestions for staying on top of latest ML/AI news? Any journals you would recommend?" }, { "id": 1667790, "postDate": "2022-01-29T00:15:40.923Z", "content": "Thank you very much for sharing! It inspired me a lot and I'm going to follow your lead.
", "rawMarkdown": "Thank you very much for sharing! It inspired me a lot and I'm going to follow your lead." }, { "id": 1666206, "postDate": "2022-01-27T14:49:44.723Z", "content": "thx for sharing, learned a lot
", "rawMarkdown": "thx for sharing, learned a lot" }, { "id": 1663822, "postDate": "2022-01-25T13:17:33.520Z", "content": "Hi Remek,
\nI just stumbled upon your post and wanted to drop a quick hello.
\nI admire your dedication and grit. Thanks for sharing your insights!
Thanks for sharing, I read your implementations of yolox and yolor and these are all very helpful!!! I wanna ask if you write your own code (without copying) and if so how? I'm quite a newbie there
", "rawMarkdown": "Thanks for sharing, I read your implementations of yolox and yolor and these are all very helpful!!! I wanna ask if you write your own code (without copying) and if so how? I'm quite a newbie there", "replies": [ { "id": 1683921, "postDate": "2022-02-10T06:23:30.817Z", "content": "I don't know about Remek's notebooks specifically, but many times you can just fork a preexisting notebook and play around with it a lot. Try different calculations, modify train/test split, etc. In the process of doing so, you will gain a lot of new ideas and build on top of the framework already laid out for you. You may even figure out a winning approach! Research really is an iterative process rather than a one time, completely from scratch implementation.
\nI also agree with Remek's point about learning the basics well and practicing a lot. The less time you have to spend figuring out how to code, the more time you have to think about what to code.
", "rawMarkdown": "I don't know about Remek's notebooks specifically, but many times you can just fork a preexisting notebook and play around with it a lot. Try different calculations, modify train/test split, etc. In the process of doing so, you will gain a lot of new ideas and build on top of the framework already laid out for you. You may even figure out a winning approach! Research really is an iterative process rather than a one time, completely from scratch implementation.\n\nI also agree with Remek's point about learning the basics well and practicing a lot. The less time you have to spend figuring out how to code, the more time you have to think about what to code." } ] }, { "id": 1650128, "postDate": "2022-01-14T20:36:18.707Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 1653786, "postDate": "2022-01-17T21:25:03.100Z", "content": "Many thanks for your great tips
", "rawMarkdown": " Many thanks for your great tips", "votes": 1 }, { "id": 1651861, "postDate": "2022-01-16T06:06:57.853Z", "content": "Thank you for sharing!!!
", "rawMarkdown": "Thank you for sharing!!!", "votes": 1 }, { "id": 1647388, "postDate": "2022-01-12T14:56:10.983Z", "content": "Great work! Thanks for sharing!
", "rawMarkdown": "Great work! Thanks for sharing!", "votes": 1 }, { "id": 1643612, "postDate": "2022-01-09T15:17:20.767Z", "content": "Thanks for sharing.
", "rawMarkdown": "Thanks for sharing.", "votes": 1 }, { "id": 1643598, "postDate": "2022-01-09T15:09:39.533Z", "content": "Thank you for sharing
", "rawMarkdown": "Thank you for sharing", "votes": 1 }, { "id": 1642563, "postDate": "2022-01-08T13:56:28.877Z", "content": "thank for this explanation
", "rawMarkdown": "thank for this explanation", "votes": 1 }, { "id": 1639409, "postDate": "2022-01-05T15:59:52.920Z", "content": "Very interesting, thanks for sharing!
", "rawMarkdown": "Very interesting, thanks for sharing!", "votes": 1 }, { "id": 1637339, "postDate": "2022-01-03T19:17:34.210Z", "content": "great! thank you for sharing this!
", "rawMarkdown": "great! thank you for sharing this!", "votes": 1 }, { "id": 1636688, "postDate": "2022-01-03T07:01:34.183Z", "content": "Thank you for sharing…!
", "rawMarkdown": "Thank you for sharing...!", "votes": 1 }, { "id": 1636760, "postDate": "2022-01-03T08:48:15.187Z", "content": "thanks for sharing such a information
", "rawMarkdown": "thanks for sharing such a information" }, { "id": 1856117, "postDate": "2022-07-15T06:20:28.073Z", "content": "Thanks for sharing this great tip.
", "rawMarkdown": "Thanks for sharing this great tip." }, { "id": 1685604, "postDate": "2022-02-11T12:24:30.707Z", "content": "Thanks for sharing…. @remekkinas
", "rawMarkdown": "Thanks for sharing.... @remekkinas " }, { "id": 1682134, "postDate": "2022-02-09T00:19:44.250Z", "content": "Thank you !
", "rawMarkdown": "Thank you !" }, { "id": 1669756, "postDate": "2022-01-30T22:58:31.833Z", "content": "Thank you for the awesome tips!
", "rawMarkdown": "Thank you for the awesome tips! " }, { "id": 1669382, "postDate": "2022-01-30T16:09:34.807Z", "content": "Thanks for the tips! :)
", "rawMarkdown": "Thanks for the tips! :)" }, { "id": 1666691, "postDate": "2022-01-28T02:03:38.283Z", "content": "Thank you for sharing!!!
", "rawMarkdown": "Thank you for sharing!!!\n" }, { "id": 1662966, "postDate": "2022-01-24T17:22:47.170Z", "content": "Thanks for sharing these great tips!
", "rawMarkdown": "Thanks for sharing these great tips!" }, { "id": 1654013, "postDate": "2022-01-18T04:01:36.903Z", "content": "Inspired me a lot, thanks
", "rawMarkdown": "Inspired me a lot, thanks" } ], "comments": [ { "id": 1637011, "author_name": "hengck23", "author_url": "", "post_date": "2022-01-03T13:23:26.220000", "content": "i can add one more point:
\nfor example, one may note that there are more FPs at the bottom of the images. These are FP that are similar partial starfish that are out of the image frame. But there is a trick. Because we are dealing with video, the starfish will move from the center of the image to the bottom edge of the image. i.e. there don't suddenly appears at bottom of images.
\nso if you have low confidence detection of \"partial starfish\" at the bottom of image, they are more likely to be FP if you cannot track from center of the image from previous frames.
\n(this is a heuristics method. if you want ML solution, you can add temporal and coordinate features to your detection/tracking model)
\nmany or tricks can be easily discovered by observations
", "votes": 17, "replies": [] }, { "id": 1633960, "author_name": "Shon Meckfessel", "author_url": "", "post_date": "2021-12-31T08:12:45.377000", "content": "Thank you for this advice! So many sites claim to be a means to community, but don't invite it in any ways beyond opening a forum. Your post really helps me understand how central humbly learning from other Kagglers is at every step. Thank you!
", "votes": 3, "replies": [ { "id": 1634131, "author_name": "Remek Kinas", "author_url": "", "post_date": "2021-12-31T11:17:51.590000", "content": "Thank you @shonmeckfessel for your feedback. I really appreciate that. I'm not the best. I am not even very good, but … I try to share my works and improve them to learn new things overnight. And I like to have fun. That's why I usually treat the competition as entertainment, not stress and a race for $$$. For me, the number one goal is to have fun and develop myself.
", "votes": 2, "replies": [] } ] }, { "id": 1646361, "author_name": "Siddhant Mahalle", "author_url": "", "post_date": "2022-01-11T16:54:47.187000", "content": "This is very useful.
", "votes": 1, "replies": [ { "id": 1650184, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-14T21:48:28.467000", "content": "Thank you! Have a nic experience with Kaggle and supporting people! This is improtant here. I have to write it as a next lessons learned. Competing is important but … having people supporting is key.
", "votes": 0, "replies": [] } ] }, { "id": 1645686, "author_name": "DO_NOT_FORGET_ME", "author_url": "", "post_date": "2022-01-11T06:22:46.673000", "content": "thanks for sharing. That's very useful for me. The sharing is very inspiring.
", "votes": 1, "replies": [] }, { "id": 1645192, "author_name": "MikolajMichaluk", "author_url": "", "post_date": "2022-01-10T18:42:46.607000", "content": "Quite useful!
", "votes": 1, "replies": [] }, { "id": 1644941, "author_name": "superpomelo", "author_url": "", "post_date": "2022-01-10T16:09:31.937000", "content": "Thanks for sharing your experience.Hope I can profit from it!!!👍
", "votes": 1, "replies": [] }, { "id": 1644695, "author_name": "Fredrik Larsson", "author_url": "", "post_date": "2022-01-10T12:18:30.753000", "content": "Thanks! Many great tips for any beginner in machine learning.
", "votes": 1, "replies": [] }, { "id": 1644572, "author_name": "Lukasz Borecki", "author_url": "", "post_date": "2022-01-10T10:55:26.820000", "content": "You can add another:
\nOne tip that wasn't added:-
\nWhile keeping to the above listed suggestions, it should be uppermost in our minds that participation in competitions is just to gain valuable experience.
\nThe objective is not winning or getting a medal.
Inciteful @remekkinas
\nThank you for sharing!
Good points
", "votes": 1, "replies": [] }, { "id": 1643333, "author_name": "Abhijeet Deshmukh", "author_url": "", "post_date": "2022-01-09T10:20:50.700000", "content": "your words are going to build our career
", "votes": 1, "replies": [] }, { "id": 1643233, "author_name": "Koshikawa54", "author_url": "", "post_date": "2022-01-09T07:59:18.583000", "content": "Thanks for sharing. This is very useful info.
\nI'm a kaggle beginner, so I can't read the source code of object detection (e.g. YOLOX) perfectly.
\nBut, I hope I can read the source code of various object detection methods someday.
Thank you so much for sharing. I'm a beginner in ML. Would you like to give me some advice?
", "votes": 1, "replies": [] }, { "id": 1642983, "author_name": "Ugo Amadi", "author_url": "", "post_date": "2022-01-08T21:14:08.753000", "content": "Thanks for sharing! As a newbie this post is very helpful. I will commit to practice even more, maybe one day I can be be as confident and reach your pace.
", "votes": 1, "replies": [] }, { "id": 1642828, "author_name": "Binaya Bajgain", "author_url": "", "post_date": "2022-01-08T18:04:31.607000", "content": "I can relate to one of the points, solution is simple than you think. I tend to take unnecessary stress trying to write a solution and making it complicated.
", "votes": 1, "replies": [] }, { "id": 1642610, "author_name": "YvonneF", "author_url": "", "post_date": "2022-01-08T14:29:53.857000", "content": "@remekkinas Thank you so much for sharing your 2021. It is a great list! I do have one question. Can you share where you \"Stay up to date with SoTA … and ML / AI news …\" outside of kaggle. Thank you.
", "votes": 1, "replies": [] }, { "id": 1642423, "author_name": "Souvik Mandal", "author_url": "", "post_date": "2022-01-08T10:55:20.587000", "content": "Great post. Hope I can do a transfer learning from this ? :)
", "votes": 1, "replies": [] }, { "id": 1642328, "author_name": "curryall", "author_url": "", "post_date": "2022-01-08T09:17:43.233000", "content": "Thanks for sharing your experience，I hope I can catch up with your pace ！
", "votes": 1, "replies": [ { "id": 1642329, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-08T09:20:38.237000", "content": "I am sure you are able to do this …. just spend as much time as possible on Kaggle.
", "votes": 0, "replies": [] } ] }, { "id": 1642285, "author_name": "Prashant Pathak", "author_url": "", "post_date": "2022-01-08T08:41:02.333000", "content": "This one is awesome !!!
", "votes": 1, "replies": [] }, { "id": 1641970, "author_name": "Lauren Hernandez", "author_url": "", "post_date": "2022-01-07T21:53:38.647000", "content": "Thank you for sharing! As a newbie, this is incredibly helpful!
", "votes": 1, "replies": [] }, { "id": 1641804, "author_name": "SHINIGAMI", "author_url": "", "post_date": "2022-01-07T18:07:26.320000", "content": "Practice, consistency, and hard work are three essential things for improving. Keep growing @remekkinas
", "votes": 1, "replies": [ { "id": 1642330, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-08T09:20:51.963000", "content": "Thank you very much!
", "votes": 0, "replies": [] } ] }, { "id": 1641335, "author_name": "adachione", "author_url": "", "post_date": "2022-01-07T10:43:12.390000", "content": "Respect!!!
", "votes": 1, "replies": [] }, { "id": 1640642, "author_name": "Saransh Gupta", "author_url": "", "post_date": "2022-01-06T16:59:53.917000", "content": "Thank You for the advice! In the post you talked about looking for a mentor, can you please describe, how should i look for a mentor?
", "votes": 1, "replies": [ { "id": 1642332, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-08T09:22:50.753000", "content": "To be honest … the best mentor is …… Kaggle …. I was trying to find somebody but unfortunately this is really hard (expensive or people do not want to help …. or … I have not found such person - read I am not good at looging for people).
", "votes": 0, "replies": [] } ] }, { "id": 1640420, "author_name": "ysss", "author_url": "", "post_date": "2022-01-06T13:20:53.747000", "content": "So inspiring! Awesome 👍
", "votes": 1, "replies": [] }, { "id": 1640301, "author_name": "Tadej Sitar", "author_url": "", "post_date": "2022-01-06T11:07:50.130000", "content": "This is so true. I started out with reading a lot of books and online courses, but when faced with a real problem I was stuck. No clue on how to proceed and where to even begin. Now I need to face my fears and accept that only practice makes perfect, if perfect is ever achievable, and just dig in and write code every day, learn from my mistakes and do better the next day. So thank you @remekkinas for the post and inspiration.
", "votes": 1, "replies": [ { "id": 1642334, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-08T09:25:06.727000", "content": "@tadejsitar thank you for your feedback! I have the same opinion … reading book is knowledge … but knowledge is not enough. To solve problems, we must have skills and practical experience.
", "votes": 1, "replies": [] } ] }, { "id": 1640175, "author_name": "AravindThota7", "author_url": "", "post_date": "2022-01-06T08:55:12.650000", "content": "Thanks for Sharing, It's very informative.
", "votes": 1, "replies": [] }, { "id": 1640028, "author_name": "WooHanYoung", "author_url": "", "post_date": "2022-01-06T06:04:26.630000", "content": "i respect you!!
", "votes": 1, "replies": [] }, { "id": 1639973, "author_name": "SandhyaKrishnan02", "author_url": "", "post_date": "2022-01-06T05:10:39.480000", "content": "Inspiring post @remekkinas!!!
\nYes, we can grow in Kaggle by practice, practice, and more practice… share our knowledge and learn from clever solutions!!!
", "votes": 1, "replies": [ { "id": 1640045, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-06T06:16:56.503000", "content": "Yes! Thank you for your comment!
", "votes": 0, "replies": [] } ] }, { "id": 1639933, "author_name": "", "author_url": "", "post_date": "2022-01-06T04:24:15.447000", "content": "", "votes": 1, "replies": [] }, { "id": 1639870, "author_name": "", "author_url": "", "post_date": "2022-01-06T02:23:36.677000", "content": "", "votes": 1, "replies": [] }, { "id": 1639838, "author_name": "", "author_url": "", "post_date": "2022-01-06T00:39:35.733000", "content": "", "votes": 1, "replies": [ { "id": 1640393, "author_name": "", "author_url": "", "post_date": "2022-01-06T12:54:49.353000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1639091, "author_name": "", "author_url": "", "post_date": "2022-01-05T11:15:15.670000", "content": "", "votes": 1, "replies": [] }, { "id": 1638746, "author_name": "", "author_url": "", "post_date": "2022-01-05T01:42:44.820000", "content": "", "votes": 1, "replies": [] }, { "id": 1638551, "author_name": "", "author_url": "", "post_date": "2022-01-04T20:21:10.480000", "content": "", "votes": 1, "replies": [] }, { "id": 1638270, "author_name": "", "author_url": "", "post_date": "2022-01-04T15:23:53.150000", "content": "", "votes": 1, "replies": [] }, { "id": 1638017, "author_name": "", "author_url": "", "post_date": "2022-01-04T11:27:35.857000", "content": "", "votes": 1, "replies": [] }, { "id": 1637717, "author_name": "", "author_url": "", "post_date": "2022-01-04T06:27:09.977000", "content": "", "votes": 1, "replies": [] }, { "id": 1637205, "author_name": "", "author_url": "", "post_date": "2022-01-03T16:48:26.140000", "content": "", "votes": 1, "replies": [] }, { "id": 1637156, "author_name": "", "author_url": "", "post_date": "2022-01-03T16:06:56.043000", "content": "", "votes": 1, "replies": [] }, { "id": 1637025, "author_name": "", "author_url": "", "post_date": "2022-01-03T13:41:11.037000", "content": "", "votes": 1, "replies": [ { "id": 1637052, "author_name": "", "author_url": "", "post_date": "2022-01-03T14:01:55.493000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1636459, "author_name": "", "author_url": "", "post_date": "2022-01-02T22:59:36.430000", "content": "", "votes": 1, "replies": [ { "id": 1636654, "author_name": "", "author_url": "", "post_date": "2022-01-03T06:33:22.417000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 1635960, "author_name": "", "author_url": "", "post_date": "2022-01-02T12:54:50.050000", "content": "", "votes": 1, "replies": [ { "id": 1635970, "author_name": "", "author_url": "", "post_date": "2022-01-02T13:03:43.597000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 1635945, "author_name": "", "author_url": "", "post_date": "2022-01-02T12:21:28.740000", "content": "", "votes": 1, "replies": [ { "id": 1635946, "author_name": "", "author_url": "", "post_date": "2022-01-02T12:24:47.563000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 1635833, "author_name": "", "author_url": "", "post_date": "2022-01-02T08:46:47.067000", "content": "", "votes": 1, "replies": [ { "id": 1636655, "author_name": "", "author_url": "", "post_date": "2022-01-03T06:33:51.643000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 1635796, "author_name": "", "author_url": "", "post_date": "2022-01-02T07:48:35.833000", "content": "", "votes": 1, "replies": [ { "id": 1635800, "author_name": "", "author_url": "", "post_date": "2022-01-02T07:52:43.597000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1635741, "author_name": "", "author_url": "", "post_date": "2022-01-02T06:34:32.137000", "content": "", "votes": 1, "replies": [ { "id": 1635805, "author_name": "", "author_url": "", "post_date": "2022-01-02T07:57:26.263000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1634836, "author_name": "", "author_url": "", "post_date": "2022-01-01T08:36:14.930000", "content": "", "votes": 1, "replies": [ { "id": 1634849, "author_name": "", "author_url": "", "post_date": "2022-01-01T08:58:29.010000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 1633670, "author_name": "", "author_url": "", "post_date": "2021-12-31T01:15:08.837000", "content": "", "votes": 1, "replies": [ { "id": 1634118, "author_name": "", "author_url": "", "post_date": "2021-12-31T11:11:19.313000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1633560, "author_name": "", "author_url": "", "post_date": "2021-12-30T22:44:02.193000", "content": "", "votes": 1, "replies": [ { "id": 1634125, "author_name": "", "author_url": "", "post_date": "2021-12-31T11:14:19.460000", "content": "", "votes": 3, "replies": [] } ] }, { "id": 1643241, "author_name": "", "author_url": "", "post_date": "2022-01-09T08:20:08.683000", "content": "", "votes": 2, "replies": [] }, { "id": 1639192, "author_name": "", "author_url": "", "post_date": "2022-01-05T13:02:43.557000", "content": "", "votes": 2, "replies": [] }, { "id": 1638399, "author_name": "", "author_url": "", "post_date": "2022-01-04T17:31:25.323000", "content": "", "votes": 2, "replies": [] }, { "id": 2309898, "author_name": "", "author_url": "", "post_date": "2023-06-20T01:00:14.990000", "content": "", "votes": 0, "replies": [] }, { "id": 1748828, "author_name": "", "author_url": "", "post_date": "2022-04-08T03:28:50.417000", "content": "", "votes": 0, "replies": [] }, { "id": 1722534, "author_name": "", "author_url": "", "post_date": "2022-03-14T16:18:37.247000", "content": "", "votes": 0, "replies": [] }, { "id": 1704858, "author_name": "", "author_url": "", "post_date": "2022-02-25T22:51:07.877000", "content": "", "votes": 0, "replies": [ { "id": 1704859, "author_name": "", "author_url": "", "post_date": "2022-02-25T22:54:13.693000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 1683905, "author_name": "", "author_url": "", "post_date": "2022-02-10T06:14:32.017000", "content": "", "votes": 0, "replies": [] }, { "id": 1667790, "author_name": "", "author_url": "", "post_date": "2022-01-29T00:15:40.923000", "content": "", "votes": 0, "replies": [] }, { "id": 1666206, "author_name": "", "author_url": "", "post_date": "2022-01-27T14:49:44.723000", "content": "", "votes": 0, "replies": [] }, { "id": 1663822, "author_name": "", "author_url": "", "post_date": "2022-01-25T13:17:33.520000", "content": "", "votes": 0, "replies": [] }, { "id": 1663632, "author_name": "", "author_url": "", "post_date": "2022-01-25T09:38:05.570000", "content": "", "votes": 0, "replies": [ { "id": 1683921, "author_name": "", "author_url": "", "post_date": "2022-02-10T06:23:30.817000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1650128, "author_name": "", "author_url": "", "post_date": "2022-01-14T20:36:18.707000", "content": "", "votes": 1, "replies": [] }, { "id": 1653786, "author_name": "", "author_url": "", "post_date": "2022-01-17T21:25:03.100000", "content": "", "votes": 1, "replies": [] }, { "id": 1651861, "author_name": "", "author_url": "", "post_date": "2022-01-16T06:06:57.853000", "content": "", "votes": 1, "replies": [] }, { "id": 1647388, "author_name": "", "author_url": "", "post_date": "2022-01-12T14:56:10.983000", "content": "", "votes": 1, "replies": [] }, { "id": 1643612, "author_name": "", "author_url": "", "post_date": "2022-01-09T15:17:20.767000", "content": "", "votes": 1, "replies": [] }, { "id": 1643598, "author_name": "", "author_url": "", "post_date": "2022-01-09T15:09:39.533000", "content": "", "votes": 1, "replies": [] }, { "id": 1642563, "author_name": "", "author_url": "", "post_date": "2022-01-08T13:56:28.877000", "content": "", "votes": 1, "replies": [] }, { "id": 1639409, "author_name": "", "author_url": "", "post_date": "2022-01-05T15:59:52.920000", "content": "", "votes": 1, "replies": [] }, { "id": 1637339, "author_name": "", "author_url": "", "post_date": "2022-01-03T19:17:34.210000", "content": "", "votes": 1, "replies": [] }, { "id": 1636688, "author_name": "", "author_url": "", "post_date": "2022-01-03T07:01:34.183000", "content": "", "votes": 1, "replies": [] }, { "id": 1636760, "author_name": "", "author_url": "", "post_date": "2022-01-03T08:48:15.187000", "content": "", "votes": 0, "replies": [] }, { "id": 1856117, "author_name": "", "author_url": "", "post_date": "2022-07-15T06:20:28.073000", "content": "", "votes": 0, "replies": [] }, { "id": 1685604, "author_name": "", "author_url": "", "post_date": "2022-02-11T12:24:30.707000", "content": "", "votes": 0, "replies": [] }, { "id": 1682134, "author_name": "", "author_url": "", "post_date": "2022-02-09T00:19:44.250000", "content": "", "votes": 0, "replies": [] }, { "id": 1669756, "author_name": "", "author_url": "", "post_date": "2022-01-30T22:58:31.833000", "content": "", "votes": 0, "replies": [] }, { "id": 1669382, "author_name": "", "author_url": "", "post_date": "2022-01-30T16:09:34.807000", "content": "", "votes": 0, "replies": [] }, { "id": 1666691, "author_name": "", "author_url": "", "post_date": "2022-01-28T02:03:38.283000", "content": "", "votes": 0, "replies": [] }, { "id": 1662966, "author_name": "", "author_url": "", "post_date": "2022-01-24T17:22:47.170000", "content": "", "votes": 0, "replies": [] }, { "id": 1654013, "author_name": "", "author_url": "", "post_date": "2022-01-18T04:01:36.903000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1632877": "This article was published on another Kaggle forum, but I decided to share it with you here as well. These are my lessons learned from various Kaggle competitions. My point of view. Maybe some of you will find it interesting, useful? Who knows?\n\nI remember the day I joined Kaggle ... I didn't know what I was doing .... I was looking for help, for explanations ... Everything seemed so complicated to me. And when I looked at hundreds of lines of Kagglers code ... I couldn't believe that I would write such things one day.\n\n- Practice, practice, practice. Practicing make you perfect. Minimize talking about AI ... maximize practice. My experience grows with the number of lines of code written (it's thousands, thousands ... lines and hundreds of notebooks - I remember one competition where I generated over 60 different notebooks).\n- Data, data ... data ... 90% ... 9% model, 1% tuning - before you move on you must know everything about dataset. It is a broad topic ... it has many elements ... but it is the basis.\n- The solutions are usually simpler than we think.\n- I like to be surprised by clever solutions proposed by smarter than myself ... it's such a nice feeling when someone breaks your head ... oh, there are many of them on Kaggle ... respect!\n- I love the moments of Kaggle breakthrough ... when the whole world stands still for a few days, no one improves the result, then someone shares an insight, a solution and the whole world moves forward ... it shows what power lies in cooperation and sharing. It is such behavior that builds progress and innovation. Breakthroughs like this are often experienced at Kaggle competitions ... valuable experience.\n- Usually the best answer is \"I don't know\" ... we have to check it and conduct many experiments to prove the thesis. If someone, without conducting an experiment, says that \"it's like this\" then 98% is wrong.\n- I like the moment when I make a submission on Kaggle and ... wait for the results. I learned not to build expectations ... so often I greeted the \"goose\" to see later that the solution is much worse than I thought ....\n- Stay up to date with SoTA ... and ML / AI news ...\n- Learn the basics - master Python ... numpy, pandas. Then Keras / Tensorflow or / and Pytorch ... classification (implement some basic architectures from scratch), then add a head to it and make a simple detector .... and create, experiment and ... have fun!\n- Talk to people ... with enthusiasts ... on Kaggle I met people from all over the world ... France, Germany, Japan, China, Korea, USA, Turkey .... Poland ... we are still trying to figure out how to \"push\" the limits \", how to jump over another 0.001 in the result ... Sometimes it is just one hyperparameter... other times hundreds lines of code ...\n- look for a mentor .... someone from whom you will be able to learn how to make a good solution even better .... \n- if you want to develop, you MUST share ... your knowledge and solutions ...\n- is it possible to be in TOP100 kagglers in one year ... yes it is ... :) \n- I often see that people optimize model at the beggining .... Why? How to optimize something which is far from good model? How to optimize unstable system. As a result, they get worse results than if they ran the model on the default parameters. This is just last part of the game. Fight for the best model without tuning. Do this one last.\n- I often see EDA .... without \"so what\" .... just ... good looking charts ... absolutely oudstanding charts .... but ... there is no conclusion ... \"so what\" ... \n- We are often tempted to use what we already know. Since I feel good in neural networks, I start with them again. However, this is a comfort zone. Do something completely different. Learn a new thing. Expand your set of available tools.\n- Why is my notebook rated so low? Why does it only have 20 votes? Check the number of forks. You will discover its true value and usefulness. Sometimes my notebook only has 20-30 votes or it lacks a few points to gold ... I look at the number of forks and I can't believe ...\n- Stay focused .... I know that daily there is a lot new ideas in competition (or your project) ... stay focused ... “If you chase two rabbits...You will not catch either one.” Luckily Kaggle provide only 5 submission daily ... this limitation is for thinking ... not Kaggle's malice.\n\nI appereciate all of you who help me grow here. Thank you!", "1637011": "i can add one more point:\n\n- always print out / draw your prediction results and errors and observe\n\nfor example, one may note that there are more FPs at the bottom of the images. These are FP that are similar partial starfish that are out of the image frame. But there is a trick. Because we are dealing with video, the starfish will move from the center of the image to the bottom edge of the image. i.e. there don't suddenly appears at bottom of images.\n\nso if you have low confidence detection of \"partial starfish\" at the bottom of image, they are more likely to be FP if you cannot track from center of the image from previous frames.\n\n(this is a heuristics method. if you want ML solution, you can add temporal and coordinate features to your detection/tracking model)\n\nmany or tricks can be easily discovered by observations", "1633960": "Thank you for this advice! So many sites claim to be a means to community, but don't invite it in any ways beyond opening a forum. Your post really helps me understand how central humbly learning from other Kagglers is at every step. Thank you!", "1646361": "This is very useful.", "1645686": "thanks for sharing. That's very useful for me. The sharing is very inspiring.", "1645192": "Quite useful!", "1644941": "Thanks for sharing your experience.Hope I can profit from it!!!👍", "1644695": "Thanks! Many great tips for any beginner in machine learning.", "1644572": "You can add another:\n\n1. Check if the ppl are getting banned on Product Feedback, if not then edit your notebook! :)", "1643773": "One tip that wasn't added:-\nWhile keeping to the above listed suggestions, it should be uppermost in our minds that participation in competitions is just to gain valuable experience.\nThe objective is not winning or getting a medal.", "1643613": "Inciteful @remekkinas\nThank you for sharing!", "1643508": "Good points ", "1643333": "your words are going to build our career ", "1643233": "Thanks for sharing. This is very useful info.\nI'm a kaggle beginner, so I can't read the source code of object detection (e.g. YOLOX) perfectly.\nBut, I hope I can read the source code of various object detection methods someday.", "1643188": "Thank you so much for sharing. I'm a beginner in ML. Would you like to give me some advice?", "1642983": "Thanks for sharing! As a newbie this post is very helpful. I will commit to practice even more, maybe one day I can be be as confident and reach your pace. ", "1642828": "I can relate to one of the points, solution is simple than you think. I tend to take unnecessary stress trying to write a solution and making it complicated. ", "1642610": "@remekkinas Thank you so much for sharing your 2021. It is a great list! I do have one question. Can you share where you \"Stay up to date with SoTA … and ML / AI news …\" outside of kaggle. Thank you.", "1642423": "Great post. Hope I can do a transfer learning from this ? :)", "1642328": "Thanks for sharing your experience，I hope I can catch up with your pace ！", "1642285": "This one is awesome !!!", "1641970": "Thank you for sharing! As a newbie, this is incredibly helpful!", "1641804": "Practice, consistency, and hard work are three essential things for improving. Keep growing @remekkinas ", "1641335": "Respect!!!", "1640642": "Thank You for the advice! In the post you talked about looking for a mentor, can you please describe, how should i look for a mentor?", "1640420": "So inspiring! Awesome 👍", "1640301": "This is so true. I started out with reading a lot of books and online courses, but when faced with a real problem I was stuck. No clue on how to proceed and where to even begin. Now I need to face my fears and accept that only practice makes perfect, *if perfect is ever achievable*, and just dig in and write code every day, learn from my mistakes and do better the next day. So thank you @remekkinas for the post and inspiration.", "1640175": "Thanks for Sharing, It's very informative.", "1640028": "i respect you!!", "1639973": "Inspiring post @remekkinas!!! \n\nYes, we can grow in Kaggle by practice, practice, and more practice... share our knowledge and learn from clever solutions!!!", "1639933": "thank you for sharing your thought. it make me on fire haha", "1639870": "This one is aweome !!!", "1639838": "Thank you for the insightful post! As someone who is just starting out this is great info. My hope is to make it to Master by the end of the year. Hopefully, I can look back on this comment and be happy with my progress. And that will be in part thanks to your insight!", "1639091": "Hope to be \"niubi \" as you after 1 year , now i am just a little little rookie .", "1638746": "Respect!!!!!", "1638551": "Thnaks for sharing", "1638270": "There are many points which I know and many which I don't know. Thank you for sharing it ", "1638017": "This is what a New Year motivation i needed!! i will try to get close enough to these goals!! THANKS A TON👍", "1637717": "Respect!!!", "1637205": "Amazing 🔥🔥", "1637156": "Thanks for the insights, great motivation!", "1637025": "Really inspiring", "1636459": "This post gives me the motivation to push forward my Kaggle journey. Thank you @remekkinas ", "1635960": "Thank you for sharing this insightful article! It is really helpful.", "1635945": "Thank you for your very helpful insight", "1635833": "Hello, thank you for this article with minor \"best practices\". I only beginner in kaggle and this article help me understand what i need for get better result :)", "1635796": "Thank you for this insightful advice. Any tips // specific resources for staying up-to-date on ML // AI news?\n", "1635741": "Great article. Very valuable for newbies.", "1634836": "It is precisely the kind of post I wanted to start my new year reading. Thanks for sharing. It is pretty informative!", "1633670": "Totally agreed. I have just joined recently as well. but have learnt a lot since then. Thanks for all people who willing to teach and share idea here. ", "1633560": "Thank you so much for sharing your know-hows, ideas and codes at this competition.\n\nI've started Kaggle just a week ago and never touched YOLO, \nbut now I can train my YOLOX model thanks to your posts!\n\nHave a good new year!", "1643241": "Thanks for these significant key points!", "1639192": "As a newcomer to this world of data analytics, reading your post was helpful. I really like the collaboration mindset which is seems prevalent here.", "1638399": "Very valuable for newbies. !Thank you...", "2309898": "Practice, practice and practice.", "1748828": "I'm gonna save and bookmark these thoughtful words and check it from time to time as long as I'm on Kaggle ❤🔥❤🔥❤", "1722534": "Practice, practice, practice.\nStay focused\nMany thanks for your great tips", "1704858": "Thank you so much for the tips! I'm curious, what was your background when starting on Kaggle? How much coding experience did you have?", "1683905": "Great insights! Thanks for sharing. Do you have any suggestions for staying on top of latest ML/AI news? Any journals you would recommend?", "1667790": "Thank you very much for sharing! It inspired me a lot and I'm going to follow your lead.", "1666206": "thx for sharing, learned a lot", "1663822": "Hi Remek,\nI just stumbled upon your post and wanted to drop a quick hello.\nI admire your dedication and grit. Thanks for sharing your insights!", "1663632": "Thanks for sharing, I read your implementations of yolox and yolor and these are all very helpful!!! I wanna ask if you write your own code (without copying) and if so how? I'm quite a newbie there", "1650128": "", "1653786": " Many thanks for your great tips", "1651861": "Thank you for sharing!!!", "1647388": "Great work! Thanks for sharing!", "1643612": "Thanks for sharing.", "1643598": "Thank you for sharing", "1642563": "thank for this explanation", "1639409": "Very interesting, thanks for sharing!", "1637339": "great! thank you for sharing this!", "1636688": "Thank you for sharing...!", "1636760": "thanks for sharing such a information", "1856117": "Thanks for sharing this great tip.", "1685604": "Thanks for sharing.... @remekkinas ", "1682134": "Thank you !", "1669756": "Thank you for the awesome tips! ", "1669382": "Thanks for the tips! :)", "1666691": "Thank you for sharing!!!\n", "1662966": "Thanks for sharing these great tips!", "1654013": "Inspired me a lot, thanks" } }