Today I decided to introduce YoloR as a new Kagglers toy for this competition. π₯π₯π
\nFirst part introduced today Train YoloR on COTS dataset (PART 1 - TRAINING) - as easy as possible to help people start with YoloR and develop this notebook
\nThis notebook introduces YoloR on Kaggle and TensorFlow - Help Protect the Great Barrier Reef competition. It shows how to train custom object detection model (COTS dataset) using YoloR. It could be good starting point for build own custom model based on YoloR detector. Full github repository you can find here - YoloR
\nThis notebook covers:
\nThis is only demo. What does it mean? This notebook is not optimized for scoring. This is part for you. I jest provided you LAB notebook for experimentations with YoloR on Kaggle.
\nInference part - this week.
\nPLEASE CONSIDER VOTING ON MY NOTEBOOK - THIS MOTIVATES ME TO CREATE MORE AND SHARE.
", "messages": [ { "id": 1631897, "postDate": "2021-12-28T22:08:37.893Z", "content": "Today I decided to introduce YoloR as a new Kagglers toy for this competition. π₯π₯π
\nFirst part introduced today Train YoloR on COTS dataset (PART 1 - TRAINING) - as easy as possible to help people start with YoloR and develop this notebook
\nThis notebook introduces YoloR on Kaggle and TensorFlow - Help Protect the Great Barrier Reef competition. It shows how to train custom object detection model (COTS dataset) using YoloR. It could be good starting point for build own custom model based on YoloR detector. Full github repository you can find here - YoloR
\nThis notebook covers:
\nThis is only demo. What does it mean? This notebook is not optimized for scoring. This is part for you. I jest provided you LAB notebook for experimentations with YoloR on Kaggle.
\nInference part - this week.
\nPLEASE CONSIDER VOTING ON MY NOTEBOOK - THIS MOTIVATES ME TO CREATE MORE AND SHARE.
", "rawMarkdown": "Today I decided to introduce YoloR as a new Kagglers toy for this competition. π₯π₯π\n\nFirst part introduced today [Train YoloR on COTS dataset (PART 1 - TRAINING) - as easy as possible to help people start with YoloR and develop this notebook](https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-train)\n\nThis notebook introduces YoloR on Kaggle and TensorFlow - Help Protect the Great Barrier Reef competition. It shows how to train custom object detection model (COTS dataset) using YoloR. It could be good starting point for build own custom model based on YoloR detector. Full github repository you can find here - [YoloR](https://github.com/WongKinYiu/yolor)\n\nThis notebook covers:\n- COTS dataset preparation for YoloR training\n- YoloR (YoloR, MISH CUDA, pytorch_wavelets) setup\n- Pre-Trained Weights for YoloR HUB download\n- Configuration files (YoloR hyperparameters and dataset) setup\n- Weights and Biases configuration for training logging\n- YoloR training\n- YoloR inference on test images\n\nThis is only demo. What does it mean? This notebook is not optimized for scoring. This is part for you. I jest provided you LAB notebook for experimentations with YoloR on Kaggle.\n\nInference part - this week.\n\n**PLEASE CONSIDER VOTING ON MY NOTEBOOK - THIS MOTIVATES ME TO CREATE MORE AND SHARE.**", "votes": 38 }, { "id": 1660880, "postDate": "2022-01-23T04:03:42.493Z", "content": "Thank you for sharing a good notebook and a new model, YOLOR!!
", "rawMarkdown": "Thank you for sharing a good notebook and a new model, YOLOR!!", "votes": 1 }, { "id": 1648047, "postDate": "2022-01-13T04:39:23.387Z", "content": "Thanks for sharing. It will help me to start on this competition.
", "rawMarkdown": "Thanks for sharing. It will help me to start on this competition. ", "votes": 1, "replies": [ { "id": 1648093, "postDate": "2022-01-13T06:05:49.360Z", "content": "You are welcome!
", "rawMarkdown": "You are welcome! " } ] }, { "id": 1634473, "postDate": "2021-12-31T17:57:54.140Z", "content": "Great work! Thanks for sharing π₯
", "rawMarkdown": "Great work! Thanks for sharing π₯", "votes": 1, "replies": [ { "id": 1634518, "postDate": "2021-12-31T18:43:35.633Z", "content": "You are welcome! :)
", "rawMarkdown": "You are welcome! :)", "votes": 1 } ] }, { "id": 1633439, "postDate": "2021-12-30T19:04:06.947Z", "content": "I have copied dataset to my g-drive, so I make little bit change to train on colab pro.
\nhttps://github.com/flydragon2018/colab_notebooks
", "rawMarkdown": "I have copied dataset to my g-drive, so I make little bit change to train on colab pro. \n\nhttps://github.com/flydragon2018/colab_notebooks", "votes": 1 }, { "id": 1632326, "postDate": "2021-12-29T13:55:30.017Z", "content": "Great! Thank you!
", "rawMarkdown": "Great! Thank you!", "votes": 1, "replies": [ { "id": 1632367, "postDate": "2021-12-29T14:43:28.027Z", "content": "You are welcome! Experiment and let us know how YoloR performs β¦ we all benefit from this.
", "rawMarkdown": "You are welcome! Experiment and let us know how YoloR performs β¦ we all benefit from this. " } ] }, { "id": 1632146, "postDate": "2021-12-29T09:06:23.873Z", "content": "Thanks for all the teaching material you are creating @remekkinas Beginners like me would be lost without them
", "rawMarkdown": "Thanks for all the teaching material you are creating @remekkinas Beginners like me would be lost without them", "votes": 1, "replies": [ { "id": 1632216, "postDate": "2021-12-29T11:11:12.070Z", "content": "I am really happy. My goal was to give people quick and easy start in this competition and β¦ in next projects. I am trying to give people rod (experimentation tool) β¦ not fish (solution). Now you can play using different hyperparameters and datset configuration. Inference will be in 2 days (today I am traveling).
", "rawMarkdown": "I am really happy. My goal was to give people quick and easy start in this competition and ... in next projects. I am trying to give people rod (experimentation tool) ... not fish (solution). Now you can play using different hyperparameters and datset configuration. Inference will be in 2 days (today I am traveling). ", "votes": 2 } ] }, { "id": 1632145, "postDate": "2021-12-29T09:04:21.860Z", "content": "great. have you converted the inference part?
\nI finally bought colab pro today, for training experiment.
", "rawMarkdown": "great. have you converted the inference part?\n\nI finally bought colab pro today, for training experiment.", "votes": 1, "replies": [ { "id": 1632175, "postDate": "2021-12-29T09:52:07.247Z", "content": "Yes. This week I will publish it. Today I am traveling all day so no coding :)
", "rawMarkdown": "Yes. This week I will publish it. Today I am traveling all day so no coding :) " } ] }, { "id": 1632119, "postDate": "2021-12-29T08:02:32.413Z", "content": "Great work!!
", "rawMarkdown": "Great work!!", "votes": 1, "replies": [ { "id": 1632217, "postDate": "2021-12-29T11:11:21.453Z", "content": "Yhank you very much!
", "rawMarkdown": "Yhank you very much!" } ] }, { "id": 1650646, "postDate": "2022-01-15T07:44:15.703Z", "content": "YoloR updated β¦ now is above 0.52 β¦. https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-infer πͺ
", "rawMarkdown": "YoloR updated ... now is above 0.52 .... https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-infer πͺ" }, { "id": 1637802, "postDate": "2022-01-04T07:52:40.247Z", "content": "@remekkinas Great work on introducing YOLOR. I tried writing inference code for YOLOR just like in YOLOX for submission purpose but failed to deliver expected results. Can you help how to use YOLOR for inference and submission? Thanks
", "rawMarkdown": "@remekkinas Great work on introducing YOLOR. I tried writing inference code for YOLOR just like in YOLOX for submission purpose but failed to deliver expected results. Can you help how to use YOLOR for inference and submission? Thanks", "replies": [ { "id": 1648131, "postDate": "2022-01-13T07:23:55.007Z", "content": "Implemented and posted in notebook section: https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-infer
", "rawMarkdown": "Implemented and posted in notebook section: https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-infer", "votes": 1 } ] }, { "id": 1632245, "postDate": "2021-12-29T12:18:57.410Z", "content": "Great work. I have tried messing with YoloR for COTS, but I haven't gotten any good results trying to train COTS on the pre-trained model. I might have just messed up some setting or hyperparameter. Hope you can get good results
", "rawMarkdown": "Great work. I have tried messing with YoloR for COTS, but I haven't gotten any good results trying to train COTS on the pre-trained model. I might have just messed up some setting or hyperparameter. Hope you can get good results", "replies": [ { "id": 1632267, "postDate": "2021-12-29T12:42:19.470Z", "content": "Yes. I will show it in inference notebook. Please consider voting on my notebook it really motivate me to create and share more. Each time I got vote β¦ each time I get a vote I feel it makes sense to spend my time sharing my knowledge.
", "rawMarkdown": "Yes. I will show it in inference notebook. Please consider voting on my notebook it really motivate me to create and share more. Each time I got vote β¦ each time I get a vote I feel it makes sense to spend my time sharing my knowledge.", "votes": 2 }, { "id": 1633433, "postDate": "2021-12-30T18:53:13.207Z", "content": "my training result so far seems not better than yolox based on the training log.
\nI quickly look into the source code, detect.py needs to make some change in order to meet the
\nspecific env.predict() submission method.
otherwise, we can just post process the saved txt file for submission.
\nSince you will release one, I will just wait and upvote you!
Thank you for sharing a good notebook and a new model, YOLOR!!
", "votes": 1, "replies": [] }, { "id": 1648047, "author_name": "Atri Saxena", "author_url": "", "post_date": "2022-01-13T04:39:23.387000", "content": "Thanks for sharing. It will help me to start on this competition.
", "votes": 1, "replies": [ { "id": 1648093, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-13T06:05:49.360000", "content": "You are welcome!
", "votes": 0, "replies": [] } ] }, { "id": 1634473, "author_name": "Ichimaru Gin", "author_url": "", "post_date": "2021-12-31T17:57:54.140000", "content": "Great work! Thanks for sharing π₯
", "votes": 1, "replies": [ { "id": 1634518, "author_name": "Remek Kinas", "author_url": "", "post_date": "2021-12-31T18:43:35.633000", "content": "You are welcome! :)
", "votes": 1, "replies": [] } ] }, { "id": 1633439, "author_name": "dragon zhang", "author_url": "", "post_date": "2021-12-30T19:04:06.947000", "content": "I have copied dataset to my g-drive, so I make little bit change to train on colab pro.
\nhttps://github.com/flydragon2018/colab_notebooks
", "votes": 1, "replies": [] }, { "id": 1632326, "author_name": "ji411", "author_url": "", "post_date": "2021-12-29T13:55:30.017000", "content": "Great! Thank you!
", "votes": 1, "replies": [ { "id": 1632367, "author_name": "Remek Kinas", "author_url": "", "post_date": "2021-12-29T14:43:28.027000", "content": "You are welcome! Experiment and let us know how YoloR performs β¦ we all benefit from this.
", "votes": 0, "replies": [] } ] }, { "id": 1632146, "author_name": "nymfree", "author_url": "", "post_date": "2021-12-29T09:06:23.873000", "content": "Thanks for all the teaching material you are creating @remekkinas Beginners like me would be lost without them
", "votes": 1, "replies": [ { "id": 1632216, "author_name": "Remek Kinas", "author_url": "", "post_date": "2021-12-29T11:11:12.070000", "content": "I am really happy. My goal was to give people quick and easy start in this competition and β¦ in next projects. I am trying to give people rod (experimentation tool) β¦ not fish (solution). Now you can play using different hyperparameters and datset configuration. Inference will be in 2 days (today I am traveling).
", "votes": 2, "replies": [] } ] }, { "id": 1632145, "author_name": "dragon zhang", "author_url": "", "post_date": "2021-12-29T09:04:21.860000", "content": "great. have you converted the inference part?
\nI finally bought colab pro today, for training experiment.
", "votes": 1, "replies": [ { "id": 1632175, "author_name": "Remek Kinas", "author_url": "", "post_date": "2021-12-29T09:52:07.247000", "content": "Yes. This week I will publish it. Today I am traveling all day so no coding :)
", "votes": 0, "replies": [] } ] }, { "id": 1632119, "author_name": "Hyeokjoon Kwon (κΆνμ€)", "author_url": "", "post_date": "2021-12-29T08:02:32.413000", "content": "Great work!!
", "votes": 1, "replies": [ { "id": 1632217, "author_name": "Remek Kinas", "author_url": "", "post_date": "2021-12-29T11:11:21.453000", "content": "Yhank you very much!
", "votes": 0, "replies": [] } ] }, { "id": 1650646, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-15T07:44:15.703000", "content": "YoloR updated β¦ now is above 0.52 β¦. https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-infer πͺ
", "votes": 0, "replies": [] }, { "id": 1637802, "author_name": "Sanchit Vijay", "author_url": "", "post_date": "2022-01-04T07:52:40.247000", "content": "@remekkinas Great work on introducing YOLOR. I tried writing inference code for YOLOR just like in YOLOX for submission purpose but failed to deliver expected results. Can you help how to use YOLOR for inference and submission? Thanks
", "votes": 0, "replies": [ { "id": 1648131, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-13T07:23:55.007000", "content": "Implemented and posted in notebook section: https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-infer
", "votes": 1, "replies": [] } ] }, { "id": 1632245, "author_name": "Lime-Cake", "author_url": "", "post_date": "2021-12-29T12:18:57.410000", "content": "Great work. I have tried messing with YoloR for COTS, but I haven't gotten any good results trying to train COTS on the pre-trained model. I might have just messed up some setting or hyperparameter. Hope you can get good results
", "votes": 0, "replies": [ { "id": 1632267, "author_name": "Remek Kinas", "author_url": "", "post_date": "2021-12-29T12:42:19.470000", "content": "Yes. I will show it in inference notebook. Please consider voting on my notebook it really motivate me to create and share more. Each time I got vote β¦ each time I get a vote I feel it makes sense to spend my time sharing my knowledge.
", "votes": 2, "replies": [] }, { "id": 1633433, "author_name": "dragon zhang", "author_url": "", "post_date": "2021-12-30T18:53:13.207000", "content": "my training result so far seems not better than yolox based on the training log.
\nI quickly look into the source code, detect.py needs to make some change in order to meet the
\nspecific env.predict() submission method.
otherwise, we can just post process the saved txt file for submission.
\nSince you will release one, I will just wait and upvote you!