{ "id": 158371, "title": "YOLOv5 is Here", "url": "/competitions/global-wheat-detection/discussion/158371", "author_name": "OrKatz", "post_date": "2020-06-14T05:10:00.245000", "votes": 9, "comment_count": 9, "views": 0, "content": "

\"\"\nhttps://models.roboflow.ai/object-detection/yolov5

", "messages": [ { "id": 885313, "postDate": "2020-06-14T05:10:00.247Z", "content": "

\"\"\nhttps://models.roboflow.ai/object-detection/yolov5

", "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1094066%2F516cb8348a18b7212725877e373a3f51%2Fyolov5-performance.png?generation=1592111422558773&alt=media)\nhttps://models.roboflow.ai/object-detection/yolov5", "votes": 9 }, { "id": 885566, "postDate": "2020-06-14T10:18:12.283Z", "content": "

That YOLOv5 is just renamed YOLOv3. The graph seems nice but it is misleading.\nFor details, see the discussion here https://github.com/AlexeyAB/darknet/issues/5920

", "rawMarkdown": "That YOLOv5 is just renamed YOLOv3. The graph seems nice but it is misleading.\nFor details, see the discussion here [https://github.com/AlexeyAB/darknet/issues/5920](https://github.com/AlexeyAB/darknet/issues/5920)\n", "votes": 6 }, { "id": 907207, "postDate": "2020-06-29T18:59:50.623Z", "content": "

Since Yolo V5 is doing so good in this competition, it's interesting to see if anybody can make Yolo V4 to gain comparable performance .

\n\n

I think both authors of V4 and V5 have contributed a lot to Yolo V3 implementation, and both of them want to make an official repo on the new Yolo version. Even though the V4 author have criticized V5 , in the official V4 paper, he did acknowledge contribution of V5 author on Mosaic augmentation which significantly improves both V4 and V5 performance.

", "rawMarkdown": "Since Yolo V5 is doing so good in this competition, it's interesting to see if anybody can make Yolo V4 to gain comparable performance . \n\nI think both authors of V4 and V5 have contributed a lot to Yolo V3 implementation, and both of them want to make an official repo on the new Yolo version. Even though the V4 author have criticized V5 , in the official V4 paper, he did acknowledge contribution of V5 author on Mosaic augmentation which significantly improves both V4 and V5 performance.", "votes": 4, "replies": [ { "id": 921424, "postDate": "2020-07-09T09:05:16.137Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 886607, "postDate": "2020-06-15T06:48:27.270Z", "content": "

I tried it with GitHub repository, and I got 0.637 LB.

", "rawMarkdown": "I tried it with [GitHub repository](https://github.com/ultralytics/yolov5), and I got 0.637 LB.", "replies": [ { "id": 886732, "postDate": "2020-06-15T08:40:31.707Z", "content": "

Thank you for sharing. According to the discussion, YOLOV3 has 0.65 so it seems \"YOLOv5\" is on par with it.

", "rawMarkdown": "Thank you for sharing. According to the [discussion](https://www.kaggle.com/c/global-wheat-detection/discussion/150780), YOLOV3 has 0.65 so it seems \"YOLOv5\" is on par with it." }, { "id": 888117, "postDate": "2020-06-16T06:26:14.543Z", "content": "

I tried it too, and I got 0.712 LB.\nI will publish the notebook soon.

", "rawMarkdown": "I tried it too, and I got 0.712 LB.\nI will publish the notebook soon.", "votes": 2 }, { "id": 888133, "postDate": "2020-06-16T06:42:23.403Z", "content": "

Just for FYI, Karol Majek (expert in object detection) compared YOLOv5 vs YOLOv4 vs YOLOv3 and here is the result (YOLOv5 is worse than YOLOv4): https://www.youtube.com/post/UgxIk5_Jlp6BQzVbSeV4AaABCQ

", "rawMarkdown": "Just for FYI, Karol Majek (expert in object detection) compared YOLOv5 vs YOLOv4 vs YOLOv3 and here is the result (YOLOv5 is worse than YOLOv4): [https://www.youtube.com/post/UgxIk5_Jlp6BQzVbSeV4AaABCQ](https://www.youtube.com/post/UgxIk5_Jlp6BQzVbSeV4AaABCQ)" }, { "id": 891356, "postDate": "2020-06-18T05:52:48.667Z", "content": "

See this notebook\nhttps://www.kaggle.com/orkatz2/yolov5-fake-or-real-single-model-l-b-0-73

", "rawMarkdown": "See this notebook\nhttps://www.kaggle.com/orkatz2/yolov5-fake-or-real-single-model-l-b-0-73", "votes": 2 }, { "id": 891688, "postDate": "2020-06-18T11:39:18.760Z", "content": "

Wow, that is really high score.

", "rawMarkdown": "Wow, that is really high score." } ] } ], "comments": [ { "id": 885566, "author_name": "Petr Hurtik", "author_url": "", "post_date": "2020-06-14T10:18:12.283000", "content": "

That YOLOv5 is just renamed YOLOv3. The graph seems nice but it is misleading.\nFor details, see the discussion here https://github.com/AlexeyAB/darknet/issues/5920

", "votes": 6, "replies": [] }, { "id": 907207, "author_name": "Neuron Engineer", "author_url": "", "post_date": "2020-06-29T18:59:50.623000", "content": "

Since Yolo V5 is doing so good in this competition, it's interesting to see if anybody can make Yolo V4 to gain comparable performance .

\n\n

I think both authors of V4 and V5 have contributed a lot to Yolo V3 implementation, and both of them want to make an official repo on the new Yolo version. Even though the V4 author have criticized V5 , in the official V4 paper, he did acknowledge contribution of V5 author on Mosaic augmentation which significantly improves both V4 and V5 performance.

", "votes": 4, "replies": [ { "id": 921424, "author_name": "", "author_url": "", "post_date": "2020-07-09T09:05:16.137000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 886607, "author_name": "Yudai Hayashi", "author_url": "", "post_date": "2020-06-15T06:48:27.270000", "content": "

I tried it with GitHub repository, and I got 0.637 LB.

", "votes": 0, "replies": [ { "id": 886732, "author_name": "Petr Hurtik", "author_url": "", "post_date": "2020-06-15T08:40:31.707000", "content": "

Thank you for sharing. According to the discussion, YOLOV3 has 0.65 so it seems \"YOLOv5\" is on par with it.

", "votes": 0, "replies": [] }, { "id": 888117, "author_name": "OrKatz", "author_url": "", "post_date": "2020-06-16T06:26:14.543000", "content": "

I tried it too, and I got 0.712 LB.\nI will publish the notebook soon.

", "votes": 2, "replies": [] }, { "id": 888133, "author_name": "Petr Hurtik", "author_url": "", "post_date": "2020-06-16T06:42:23.403000", "content": "

Just for FYI, Karol Majek (expert in object detection) compared YOLOv5 vs YOLOv4 vs YOLOv3 and here is the result (YOLOv5 is worse than YOLOv4): https://www.youtube.com/post/UgxIk5_Jlp6BQzVbSeV4AaABCQ

", "votes": 0, "replies": [] }, { "id": 891356, "author_name": "OrKatz", "author_url": "", "post_date": "2020-06-18T05:52:48.667000", "content": "

See this notebook\nhttps://www.kaggle.com/orkatz2/yolov5-fake-or-real-single-model-l-b-0-73

", "votes": 2, "replies": [] }, { "id": 891688, "author_name": "Petr Hurtik", "author_url": "", "post_date": "2020-06-18T11:39:18.760000", "content": "

Wow, that is really high score.

", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "885313": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1094066%2F516cb8348a18b7212725877e373a3f51%2Fyolov5-performance.png?generation=1592111422558773&alt=media)\nhttps://models.roboflow.ai/object-detection/yolov5", "885566": "That YOLOv5 is just renamed YOLOv3. The graph seems nice but it is misleading.\nFor details, see the discussion here [https://github.com/AlexeyAB/darknet/issues/5920](https://github.com/AlexeyAB/darknet/issues/5920)\n", "907207": "Since Yolo V5 is doing so good in this competition, it's interesting to see if anybody can make Yolo V4 to gain comparable performance . \n\nI think both authors of V4 and V5 have contributed a lot to Yolo V3 implementation, and both of them want to make an official repo on the new Yolo version. Even though the V4 author have criticized V5 , in the official V4 paper, he did acknowledge contribution of V5 author on Mosaic augmentation which significantly improves both V4 and V5 performance.", "886607": "I tried it with [GitHub repository](https://github.com/ultralytics/yolov5), and I got 0.637 LB." } }