Hello there! I'm a beginner in ML and I just started out learning everything on Kaggle courses a few months ago. I'm really interested in joining this competition because I want to apply my ML skills and learn more about this area (object detection).
\nHowever, it seems like there is a lot of things I still have to learn about ML and object detection to even get started with the competition - and it really is overwhelming. So if anyone can guide me/advice on what would be the next best step(s) to do in order to get started/to study and learn about I'd really appreciate it 🙏
", "messages": [ { "id": 1593070, "postDate": "2021-11-23T15:56:40.547Z", "content": "Hello there! I'm a beginner in ML and I just started out learning everything on Kaggle courses a few months ago. I'm really interested in joining this competition because I want to apply my ML skills and learn more about this area (object detection).
\nHowever, it seems like there is a lot of things I still have to learn about ML and object detection to even get started with the competition - and it really is overwhelming. So if anyone can guide me/advice on what would be the next best step(s) to do in order to get started/to study and learn about I'd really appreciate it 🙏
", "rawMarkdown": "Hello there! I'm a beginner in ML and I just started out learning everything on Kaggle courses a few months ago. I'm really interested in joining this competition because I want to apply my ML skills and learn more about this area (object detection). \n\nHowever, it seems like there is a lot of things I still have to learn about ML and object detection to even get started with the competition - and it really is overwhelming. So if anyone can guide me/advice on what would be the next best step(s) to do in order to get started/to study and learn about I'd really appreciate it 🙏", "votes": 10 }, { "id": 1602161, "postDate": "2021-12-01T20:02:51.173Z", "content": "Begin with the aim to always learn. Take it one chunk at a time: understand the problem of the competition and what's being attempted to be solved with it; explore discussions and engages; learn from others through their code and approaches; and lastly, experiment, discover, and share…. this is how i started, and now I'm still discovering new ways of learning and developing on Kaggle. just dive in.
\nhope it helps
", "rawMarkdown": "Begin with the aim to always learn. Take it one chunk at a time: understand the problem of the competition and what's being attempted to be solved with it; explore discussions and engages; learn from others through their code and approaches; and lastly, experiment, discover, and share.... this is how i started, and now I'm still discovering new ways of learning and developing on Kaggle. just dive in. \n\nhope it helps", "votes": 1 }, { "id": 1601981, "postDate": "2021-12-01T17:17:31.803Z", "content": "I remember I tried to participate in VinBigData competition (my first object detection competition), and I got smashed :D
\nimho, object detection is more about figuring out how to use object detection libraries.
\nIt's hard to participate in this kind of competitions for the first time, since you haven't used one yet.
I believe, after several object detection competitions you will have strong codebase which will allow you to prototype/be one with the code in much earlier stages of competition hence you'll have more time to do experiments.
\nAlso, for some competitions you need a good hardware, - I saw people reporting that they're training their models with 1280 resolution which may be impossible due to small VRAM on most GPUs that are provided for free (e.g. Colab/Kaggle)
\nTL; DR;
\nobject detection competitions are not beginner friendly :D (maybe, that's my experience)
But for the starters, I would recommend you to check previous notebooks that are using YOLO family, for example, this notebook is using yolov5 : https://www.kaggle.com/h053473666/siim-cov19-yolov5-train
\nAnd this repository should also guide you how to setup and use yolov5 : https://github.com/ultralytics/yolov5
", "rawMarkdown": "I remember I tried to participate in VinBigData competition (my first object detection competition), and I got smashed :D\n\nimho, object detection is more about figuring out how to use object detection libraries.\nIt's hard to participate in this kind of competitions for the first time, since you haven't used one yet.\n\nI believe, after several object detection competitions you will have strong codebase which will allow you to prototype/be one with the code in much earlier stages of competition hence you'll have more time to do experiments.\n\nAlso, for some competitions you need a good hardware, - I saw people reporting that they're training their models with 1280 resolution which may be impossible due to small VRAM on most GPUs that are provided for free (e.g. Colab/Kaggle)\n\nTL; DR;\nobject detection competitions are not beginner friendly :D (maybe, that's my experience)\n\nBut for the starters, I would recommend you to check previous notebooks that are using YOLO family, for example, this notebook is using yolov5 : https://www.kaggle.com/h053473666/siim-cov19-yolov5-train\n\nAnd this repository should also guide you how to setup and use yolov5 : https://github.com/ultralytics/yolov5", "votes": 1 }, { "id": 1593157, "postDate": "2021-11-23T17:19:05.973Z", "content": "1- Start with good notebooks..like this.
\nhttps://www.kaggle.com/khanhlvg/cots-detection-w-tensorflow-object-detection-api
\nhttps://www.kaggle.com/khanhlvg/inference-using-efficientdet-d0-model-tensorflow
\n2- Try to understand every line of code there.
\n3- Try to experiment with changing the code (only one change at a time)
\n4- Track your experimentations.
\n5- Read and follow important discussions
\n6- Read and follow important notebooks
\n7- Enjoy
Sounds good! Thank you
", "rawMarkdown": "Sounds good! Thank you", "votes": 1 } ] }, { "id": 1593574, "postDate": "2021-11-24T04:25:04.253Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ], "comments": [ { "id": 1602161, "author_name": "ElenaEB", "author_url": "", "post_date": "2021-12-01T20:02:51.173000", "content": "Begin with the aim to always learn. Take it one chunk at a time: understand the problem of the competition and what's being attempted to be solved with it; explore discussions and engages; learn from others through their code and approaches; and lastly, experiment, discover, and share…. this is how i started, and now I'm still discovering new ways of learning and developing on Kaggle. just dive in.
\nhope it helps
", "votes": 1, "replies": [] }, { "id": 1601981, "author_name": "slime", "author_url": "", "post_date": "2021-12-01T17:17:31.803000", "content": "I remember I tried to participate in VinBigData competition (my first object detection competition), and I got smashed :D
\nimho, object detection is more about figuring out how to use object detection libraries.
\nIt's hard to participate in this kind of competitions for the first time, since you haven't used one yet.
I believe, after several object detection competitions you will have strong codebase which will allow you to prototype/be one with the code in much earlier stages of competition hence you'll have more time to do experiments.
\nAlso, for some competitions you need a good hardware, - I saw people reporting that they're training their models with 1280 resolution which may be impossible due to small VRAM on most GPUs that are provided for free (e.g. Colab/Kaggle)
\nTL; DR;
\nobject detection competitions are not beginner friendly :D (maybe, that's my experience)
But for the starters, I would recommend you to check previous notebooks that are using YOLO family, for example, this notebook is using yolov5 : https://www.kaggle.com/h053473666/siim-cov19-yolov5-train
\nAnd this repository should also guide you how to setup and use yolov5 : https://github.com/ultralytics/yolov5
", "votes": 1, "replies": [] }, { "id": 1593157, "author_name": "Faisal Alsrheed", "author_url": "", "post_date": "2021-11-23T17:19:05.973000", "content": "1- Start with good notebooks..like this.
\nhttps://www.kaggle.com/khanhlvg/cots-detection-w-tensorflow-object-detection-api
\nhttps://www.kaggle.com/khanhlvg/inference-using-efficientdet-d0-model-tensorflow
\n2- Try to understand every line of code there.
\n3- Try to experiment with changing the code (only one change at a time)
\n4- Track your experimentations.
\n5- Read and follow important discussions
\n6- Read and follow important notebooks
\n7- Enjoy
Sounds good! Thank you
", "votes": 1, "replies": [] } ] }, { "id": 1593574, "author_name": "", "author_url": "", "post_date": "2021-11-24T04:25:04.253000", "content": "", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "1593070": "Hello there! I'm a beginner in ML and I just started out learning everything on Kaggle courses a few months ago. I'm really interested in joining this competition because I want to apply my ML skills and learn more about this area (object detection). \n\nHowever, it seems like there is a lot of things I still have to learn about ML and object detection to even get started with the competition - and it really is overwhelming. So if anyone can guide me/advice on what would be the next best step(s) to do in order to get started/to study and learn about I'd really appreciate it 🙏", "1602161": "Begin with the aim to always learn. Take it one chunk at a time: understand the problem of the competition and what's being attempted to be solved with it; explore discussions and engages; learn from others through their code and approaches; and lastly, experiment, discover, and share.... this is how i started, and now I'm still discovering new ways of learning and developing on Kaggle. just dive in. \n\nhope it helps", "1601981": "I remember I tried to participate in VinBigData competition (my first object detection competition), and I got smashed :D\n\nimho, object detection is more about figuring out how to use object detection libraries.\nIt's hard to participate in this kind of competitions for the first time, since you haven't used one yet.\n\nI believe, after several object detection competitions you will have strong codebase which will allow you to prototype/be one with the code in much earlier stages of competition hence you'll have more time to do experiments.\n\nAlso, for some competitions you need a good hardware, - I saw people reporting that they're training their models with 1280 resolution which may be impossible due to small VRAM on most GPUs that are provided for free (e.g. Colab/Kaggle)\n\nTL; DR;\nobject detection competitions are not beginner friendly :D (maybe, that's my experience)\n\nBut for the starters, I would recommend you to check previous notebooks that are using YOLO family, for example, this notebook is using yolov5 : https://www.kaggle.com/h053473666/siim-cov19-yolov5-train\n\nAnd this repository should also guide you how to setup and use yolov5 : https://github.com/ultralytics/yolov5", "1593157": "1- Start with good notebooks..like this.\nhttps://www.kaggle.com/khanhlvg/cots-detection-w-tensorflow-object-detection-api\nhttps://www.kaggle.com/khanhlvg/inference-using-efficientdet-d0-model-tensorflow\n2- Try to understand every line of code there.\n3- Try to experiment with changing the code (only one change at a time) \n4- Track your experimentations.\n5- Read and follow important discussions\n6- Read and follow important notebooks\n7- Enjoy", "1593574": "" } }