deep_shf comments

Results 61 comments of


                                            deep_shf

ModuleNotFoundError: No module named 'utils.torch_utils'

w我把yoloV7的utils文件夹丢进去在加在模型时候报错 ModuleNotFoundError: No module named 'models'

自定义数据集训练问题

> 请问我在一个有两类目标的任务上训练，在测试集上感觉finetune后mAP比pretrain提高的精度非常有限，我训练的时候只使用了业务数据，并且训练类别改成了2，请问这个要怎么继续提升精度呢。你自己的训练集大不大的？

请教一下您是怎么把pth模型转换为ncnn的，我转完模型后出现了很大的偏差

您解决了吗

pred_distri, pred_scores = torch.cat([xi.view(feats[0].shape[0], self.no, -1) for xi in feats], 2).split((self.reg_max * 4, self.nc), 1) RuntimeError: shape '[70, 70, -1]' is invalid for input of size 164640

> using train_dual.py、yolov9.yaml the erro occur when open compute_loss ![image](https://private-user-images.githubusercontent.com/49717982/310338239-97c1b7e4-5eee-4081-bd63-152169345bf8.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.x3XWrjJM0HzKMykXmUqNT5sm4O5LP4BagFeYHYDcLEk) 请问解决了吗如何解决的

pred_distri, pred_scores = torch.cat([xi.view(feats[0].shape[0], self.no, -1) for xi in feats], 2).split((self.reg_max * 4, self.nc), 1) RuntimeError: shape '[70, 70, -1]' is invalid for input of size 164640

> I have the same problem 解决了吗

ModuleNotFoundError: No module named 'cuda'

ModuleNotFoundError: No module named 'cuda' when i run TensorRT-release-10.0/samples/python/efficientnet/infer.py how to install that libs ?

Configuration!!

python vision?

How are the qualitative detection results displayed?

> Hi! You can modify Detic's demo code available [here](https://github.com/facebookresearch/Detic/blob/main/demo.py) for visualization. great work!!! How to test the ability of open vocabulary object detection? Can you provide more details on...

[BUG/Help] 清华网盘文件下载不了，显示文件size too large

n你需要一个个的点它去下再就好了，别打包下载

我在多次训练不同类别的数据后，导致了最初的类别已经无法识别了

> 进行了模型多次训练迭代累加，导致了现在只能识别出最新训练的类别，其他的类别都无法再识别到，请问是哪里出现了问题？是否有灾难性遗忘问题？是不是没打乱顺序训练啊 ..