Zijie Guo
Zijie Guo
Sure it does. Check `README.md` ``` Train you can set parameters in cfg.py. python train.py -g [GPU_ID] -dir [Dataset direction] ... ```
`ori_img` is empty.
> > `ori_img` 是空的。 > > How can this be solved? Can you tell me the idea? I would first check all the images in the val-set. Manually. Then using...
Check your bboxes. They should not come outside of the image
> 请问你的问题如何解决的 Some of your bbox comes out of the image. Clip the box and you are good to go.
Can't really say what cause this, since you only show your final result. No other info about the thing you did. How is the loss goes? Your dataset volumn and...
> 大致了解应该是应为mAP=0导致的没有best.pt的生成 @jingtianyilong 那么请问怎么解决mAP=0的问题,是因为我使用的数据集的标签中间有空格还是其他原因吗,我的train_annotation.txt是如下格式 >  I use my own code. Not this one. So I got no clue
CUDA_VISIBLE_DEVICES=0 python3 eval_voc.py --weight_path weight/best.pt --gpu_id 0 --eval --mode det
If you update to the latest version you should see the attention block. And you can set `showatt` to `True`. Otherwise, I would suggest that you abandon the attention part.
Your settings? And your data?
Batch size 2 is too small imo. Network hardly learn anything. I normally use 16. 8 is also a good size. But this relates largely to your GPU. If 2...