yolov5_obb icon indicating copy to clipboard operation
yolov5_obb copied to clipboard

yolov5 + csl_label.(Oriented Object Detection)(Rotation Detection)(Rotated BBox)基于yolov5的旋转目标检测

Results 160 yolov5_obb issues
Sort by recently updated
recently updated
newest added

验证的图片也没有这张 为什么会出现这个错误 ![image](https://user-images.githubusercontent.com/109265747/187626513-2a23e217-dd75-4bcd-8620-b05cf420722d.png)

git上拉下项目,过完install.md,用预训练模型可以运行 python val.py --task 'val' --device 0 --save-json --batch-size 2 --data 'data/yolov5obb_demo_split.yaml' --name 'obb_demo_split' 但是自己训练的时候,使用 python train.py --device 3 报错RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory 使用项目自带的数据集,python...

val: WARNING: dataset/split_val/images/P2802__1__824___4225.png: ignoring corrupt image/label: 'dict' object has no attribute 'index' Class Images Labels P R [email protected] [email protected]:.95: 100%|██████████| 560/560 [00:34

数据集是自制的,训练过程、预测过程都能跑通,但是实际的效果存在一些比较奇怪的问题 1、召回率非常低,漏检情况特别严重 2、best.pt权重的实际效果,可能还不如 last.pt的效果 3、训练中train/val box_loss基本上不变 下面是我的各个文件 hyp.yaml&opt.yaml ![hyp opt](https://user-images.githubusercontent.com/77819387/186801634-dea0abc3-b836-406c-9529-59df26af0d0c.png) results.png ![image](https://user-images.githubusercontent.com/77819387/186804704-76e16cae-edfa-4f4a-bdc3-3c89eb553164.png) train_batch0,jpg&train_batch1.jpg ![image](https://user-images.githubusercontent.com/77819387/186804834-5629ab8c-6409-4a72-bef0-4245ab4603bf.png) train_batch2,jpg&train_batch3.jpg ![image](https://user-images.githubusercontent.com/77819387/186804893-f0aec73c-b00d-4881-a14d-d2695d0dcb75.png) val_batch0_labels.jpg&val_batch0_pred ![image](https://user-images.githubusercontent.com/77819387/186805052-0539fb61-284a-48db-97c9-38dae1d16981.png) val_batch1_labels.jpg&val_batch1_pred ![image](https://user-images.githubusercontent.com/77819387/186805126-b9a39d88-cf7e-475c-a520-96e4cde3a169.png)

我使用自己的数据集训练出现了以下问题,想请教一下: 训练数据集270张,数据集建立的是单类、密集小目标识别,进行训练,效果并不好:map一直为0 ![labels_correlogram](https://user-images.githubusercontent.com/93133444/185781635-96f58a90-705a-41ac-bed2-557ee680d45f.jpg) ![labels_xyls](https://user-images.githubusercontent.com/93133444/185781645-cf792840-9f45-4694-b334-9f9315955433.jpg) ![results](https://user-images.githubusercontent.com/93133444/185781653-209b514b-6206-4031-83da-cc427d7207e8.png) ![train_batch0](https://user-images.githubusercontent.com/93133444/185781654-2fa06d10-768d-415f-98a8-d03eecea3795.jpg) ![train_batch1](https://user-images.githubusercontent.com/93133444/185781656-159c5c48-7e0b-4177-bdc5-90c98f301d2d.jpg) ![train_batch2](https://user-images.githubusercontent.com/93133444/185781659-e4bba5ed-6f3d-4747-ab24-6644449b21b4.jpg) val_batch0_labels.jpg ![val_batch0_labels](https://user-images.githubusercontent.com/93133444/185781726-5a6e5a66-3bad-4f7c-9cc0-d221fc1e0da3.jpg) val_batch0_pred.jpg ![val_batch0_pred](https://user-images.githubusercontent.com/93133444/185781737-e419d75f-e4e4-4415-8893-bba5efba5afd.jpg) hyp: ``` lr0: 0.01 lrf: 0.2 momentum: 0.937 weight_decay: 0.0005 warmup_epochs: 150 warmup_momentum: 0.8 warmup_bias_lr: 0.1 box: 0.05...

直接从仓库中拉取代码,并用以下命令直接训练 `python train.py --weights weights/yolov5x.pt --data data/yolov5obb_demo.yaml --hyp data/hyps/obb/hyp.finetune_dota.yaml --epochs 10 --batch-size 1 --img 1024 --device 0` 训练过程中loss下降,但验证的时候指标均为0,并且显示labels也为0

Thanks for sharing your project with the community. I would like to know, how did the model calculated the rotated angle of the detected objects ? Can you please share...

![image](https://user-images.githubusercontent.com/37990327/185296246-ca353ef6-86d8-4647-bade-e87f26c93154.png) 训练时出现上图的错误,按照install.md和getstart.md执行的,不知道哪里出现错误。

我想将预测框长边和短边的颜色设置不同,该如何修改?

我想定义一个新的评价指标,比如将角度偏差定义小于10度,iou大于0.5的框定义为tp,我应该在哪里改动?