This is a collection of links to the top ranking notebooks, categorized by object detection model architecture. These introductory notebooks have been invaluable to many people (including myself) for installing / running the training and inference for this competition. Hope this might be of use to newcomers / latecomers to this competition!
\nPlease let me know if I have missed any important notebooks. To keep this list clean, the below list contains original notebooks. Notebooks forked from these / otherwise \"heavily inspired\" notebooks have not been included.
\nThis is a collection of links to the top ranking notebooks, categorized by object detection model architecture. These introductory notebooks have been invaluable to many people (including myself) for installing / running the training and inference for this competition. Hope this might be of use to newcomers / latecomers to this competition!
\nPlease let me know if I have missed any important notebooks. To keep this list clean, the below list contains original notebooks. Notebooks forked from these / otherwise \"heavily inspired\" notebooks have not been included.
\nnice summary of resources.
", "rawMarkdown": "nice summary of resources.", "votes": null }, { "id": "1697473", "postDate": "02/19/2022 16:38:54", "content": "Thank you for sharing @alexchwong you missed on the centernet, 3rd place solution was also using that, and there is also a public NB on this,
\n", "rawMarkdown": "Thank you for sharing @alexchwong you missed on the centernet, 3rd place solution was also using that, and there is also a public NB on this, \n- CenterNet2 [https://www.kaggle.com/sahilchachra/barrier-reef-centernet2-detectron2-training]", "votes": null }, { "id": "1697870", "postDate": "02/19/2022 23:16:12", "content": "Apologies for my ignorance. Post edited to include CenterNet2
\nI'd much prefer if the notebook has a proper inference pipeline including competition score. Even though there are training metrics, it would be nice to see how the model performs in the competition metrics.
\nIf team hydrogen ( @philippsinger and @ybabakhin) wouldn't mind publishing their CenterNet-based solution as a stand-alone notebook I'd definitely include theirs on this list as well.
", "rawMarkdown": "Apologies for my ignorance. Post edited to include CenterNet2\n\nI'd much prefer if the notebook has a proper inference pipeline including competition score. Even though there are training metrics, it would be nice to see how the model performs in the competition metrics.\n\nIf team hydrogen ( @philippsinger and @ybabakhin) wouldn't mind publishing their CenterNet-based solution as a stand-alone notebook I'd definitely include theirs on this list as well.", "votes": null }, { "id": "1699480", "postDate": "02/21/2022 07:58:15", "content": "here you go 😄
\ncenternet
Thanks @outrunner . Could you provide a notebook that shows how the model is trained?
", "rawMarkdown": "Thanks @outrunner . Could you provide a notebook that shows how the model is trained?", "votes": null }, { "id": "1699493", "postDate": "02/21/2022 08:08:05", "content": "no. :)
", "rawMarkdown": "no. :)", "votes": null } ], "comments": [ { "id": 1696793, "author_name": "dragonzhang", "author_url": "", "post_date": "02/19/2022 05:34:11", "content": "nice summary of resources.
", "votes": null, "replies": [] }, { "id": 1697473, "author_name": "soumya9977", "author_url": "", "post_date": "02/19/2022 16:38:54", "content": "Thank you for sharing @alexchwong you missed on the centernet, 3rd place solution was also using that, and there is also a public NB on this,
\n", "votes": null, "replies": [ { "id": 1697870, "author_name": "alexchwong", "author_url": "", "post_date": "02/19/2022 23:16:12", "content": "Apologies for my ignorance. Post edited to include CenterNet2
\nI'd much prefer if the notebook has a proper inference pipeline including competition score. Even though there are training metrics, it would be nice to see how the model performs in the competition metrics.
\nIf team hydrogen ( @philippsinger and @ybabakhin) wouldn't mind publishing their CenterNet-based solution as a stand-alone notebook I'd definitely include theirs on this list as well.
", "votes": null, "replies": [] }, { "id": 1699480, "author_name": "outrunner", "author_url": "", "post_date": "02/21/2022 07:58:15", "content": "here you go 😄
\ncenternet
Thanks @outrunner . Could you provide a notebook that shows how the model is trained?
", "votes": null, "replies": [] }, { "id": 1699493, "author_name": "outrunner", "author_url": "", "post_date": "02/21/2022 08:08:05", "content": "no. :)
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1696771": "This is a collection of links to the top ranking notebooks, categorized by object detection model architecture. These introductory notebooks have been invaluable to many people (including myself) for installing / running the training and inference for this competition. Hope this might be of use to newcomers / latecomers to this competition!\n\nPlease let me know if I have missed any important notebooks. To keep this list clean, the below list contains ***original*** notebooks. Notebooks forked from these / otherwise \"heavily inspired\" notebooks have not been included.\n\n### **YOLOv5**\n##### Github Repo:\n- https://github.com/ultralytics/yolov5\n##### Training:\n- https://www.kaggle.com/awsaf49/great-barrier-reef-yolov5-train Great-Barrier-Reef: YOLOv5 [train] 🌊\n- https://www.kaggle.com/steamedsheep/yolov5-high-resolution-training yolov5 high resolution training\n- https://www.kaggle.com/andradaolteanu/greatbarrierreef-yolo-full-guide-train-infer 🐡GreatBarrierReef: YOLO Full Guide [train+infer]\n##### Inference:\n- https://www.kaggle.com/awsaf49/great-barrier-reef-yolov5-infer Great-Barrier-Reef: YOLOv5 [infer] 🌊\n- https://www.kaggle.com/steamedsheep/yolov5-is-all-you-need Yolov5 is all you need\n- https://www.kaggle.com/andradaolteanu/greatbarrierreef-yolo-full-guide-train-infer 🐡GreatBarrierReef: YOLO Full Guide [train+infer]\n\n### **YOLOX**\n##### Github Repo:\n- https://github.com/Megvii-BaseDetection/YOLOX\n##### Training:\n- https://www.kaggle.com/remekkinas/yolox-full-training-pipeline-for-cots-dataset YoloX training pipeline COTS dataset [LB 0.507] !!\n##### Inference:\n- https://www.kaggle.com/remekkinas/yolox-inference-on-kaggle-for-cots-lb-0-507 YoloX inference on Kaggle for COTS [LB 0.507] !!! \n\n### **TensorFlow / EfficientDet**\n##### Github Repo:\n- https://github.com/tensorflow/models/tree/master/research/object_detection\n##### Training:\n- https://www.kaggle.com/khanhlvg/cots-detection-w-tensorflow-object-detection-api COTS detection w/ TensorFlow Object Detection API\n##### Inference:\n- https://www.kaggle.com/khanhlvg/inference-using-efficientdet-d0-model-tensorflow Inference using EfficientDet-D0 model (TensorFlow)\n\n### **YOLOR**\n##### Github Repo:\n- https://github.com/WongKinYiu/yolor\n##### Training:\n- https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-train YoloR [P6/W6]... one more yolo on Kaggle [TRAIN]\n##### Inference:\n- https://www.kaggle.com/remekkinas/yolor-p6-w6-one-more-yolo-on-kaggle-infer YoloR [P6/W6] ... one more yolo on Kaggle [INFER]\n\n### **Faster RCNN**\n##### Training:\n- https://www.kaggle.com/julian3833/reef-starter-torch-fasterrcnn-train-lb-0-416 🐠 Reef- Starter Torch FasterRCNN Train [LB=0.416]\n- https://www.kaggle.com/mlneo07/mmdetection-swin-transformer-fasterrcnn-training MMDetection Swin Transformer FasterRCNN [Training]\n##### Inference:\n- https://www.kaggle.com/julian3833/reef-starter-torch-fasterrcnn-infer-lb-0-416 🐠 Reef- Starter Torch FasterRCNN Infer [LB=0.416]\n- https://www.kaggle.com/mlneo07/mmdetection-swin-transfomer-frcnn-inference-0-443 MMDetection Swin Transfomer FRCNN[Inference 0.443]\n\n### **DETR**\n##### Github Repo:\n- https://github.com/facebookresearch/detr\n##### Training:\n- https://www.kaggle.com/julian3833/detr-detection-transformer-train-0-189 🐠 DETR - Detection Transformer - Train [0.189]\n##### Inference:\n- https://www.kaggle.com/julian3833/detr-detection-transformer-infer-0-189 🐠 DETR - Detection Transformer - Infer [0.189]\n\n### **RetinaNet**\n##### Github Repo:\n- https://github.com/fizyr/keras-retinanet.git\n##### Training:\n- https://www.kaggle.com/mahipalsingh/detection-using-keras-retinanet-train 🌟🐟Detection using Keras-RetinaNet [Train]\n##### Inference:\n- https://www.kaggle.com/mahipalsingh/detection-using-keras-retinanet-inference 🌟🐟Detection using Keras-RetinaNet [Inference]\n\n### **Detectron2**\n##### Github Repo:\n- https://github.com/facebookresearch/detectron2\n##### Training:\n- https://www.kaggle.com/ammarnassanalhajali/barrier-reef-detectron2-training Barrier Reef Detectron2 [Training]\n##### Inference:\n- https://www.kaggle.com/ammarnassanalhajali/barrier-reef-detectron2-inference Barrier Reef Detectron2 [Inference]\n\n### **CenterNet2**\n##### Github Repo:\n- https://github.com/xingyizhou/CenterNet2\n##### Training only:\n- https://www.kaggle.com/sahilchachra/barrier-reef-centernet2-detectron2-training Barrier Reef - CenterNet2 (Detectron2) [Training]\n##### Inference only:\n- https://www.kaggle.com/outrunner/centernet-s CenterNet S (by outrunner)\n\n### **YOLOv4**\n##### Github Repo:\n- https://github.com/AlexeyAB/darknet.git\n##### Training / Inference:\n- https://www.kaggle.com/gimarcecaml/cots-det-yolov4-darknet-install-train-infer COTS det: YOLOv4 (darknet) [install, train, infer]\n\n### **ScaledYOLOv4**\n##### Github Repo:\n- https://github.com/WongKinYiu/ScaledYOLOv4\n##### Training\n- https://www.kaggle.com/alexchwong/scaledyolov4-for-cots-training ScaledYOLOv4 for COTS [Training]\n##### Inference\n- https://www.kaggle.com/alexchwong/scaledyolov4-for-cots-inference ScaledYOLOv4 for COTS [Inference]", "1696793": "nice summary of resources.", "1697473": "Thank you for sharing @alexchwong you missed on the centernet, 3rd place solution was also using that, and there is also a public NB on this, \n- CenterNet2 [https://www.kaggle.com/sahilchachra/barrier-reef-centernet2-detectron2-training]", "1697870": "Apologies for my ignorance. Post edited to include CenterNet2\n\nI'd much prefer if the notebook has a proper inference pipeline including competition score. Even though there are training metrics, it would be nice to see how the model performs in the competition metrics.\n\nIf team hydrogen ( @philippsinger and @ybabakhin) wouldn't mind publishing their CenterNet-based solution as a stand-alone notebook I'd definitely include theirs on this list as well.", "1699480": "here you go 😄\n[centernet](https://www.kaggle.com/outrunner/centernet-s)", "1699489": "Thanks @outrunner . Could you provide a notebook that shows how the model is trained?", "1699493": "no. :)" }, "source": "meta" }