thohemp / cube_detector Goto Github PK

Hi, I was wondering if it might be possible to run your model even on the limited computing power of a smartphone. Before diving too deep into it, I wanted to get the opinion of more experienced people in the field.

If you think it is possible, is a rough estimation for fps possible ?

If you think it is not possible could you briefly elaborate on the reasons please ?

Any insights are appreciated. Thank you very much!

Sorry to bother you again.

I succeeded training with cube dataset with your help.

And I am trying to train my own dataset.
I want to detect the rotated vehicle.

This time I am little confused with the angle.

I think you use 0 to 90 angle because the cube is square. right?

But because vehicle is rectangle, I think I cannot express the angle between 0 and 90.

So I changed the angle between 0 and 180, and correct the source code to train 0 and 180 angle.

But the mAP doesn't increase at all.

Can you advice me how to train?

python detect.py --source 0 --weights m640rot.pt --imgsz 640 --conf-thres 0.75 ### when i run the code i got this error. Can u help me. Thx..

line 571, in non_max_suppression
conf_angle, j_angle = x[:, 5+nc:].max(1, keepdim=True)
RuntimeError: cannot perform reduction function max on tensor with no elements because the operation does not have an identity

Hello, I understand that the angle of the project is basically set from 0 to 90, and we have modified it to 0 to 360 and have used it well when the class num is 1 However, if the class num is more than 2, the following error occurs Do you know anything?
Thank you.

Traceback (most recent call last):
File "./yolov5/train.py", line 600, in
main(opt)
File "./yolov5/train.py", line 498, in main
train(opt.hyp, opt, device)
File "./yolov5/train.py", line 363, in train
compute_loss=compute_loss)
File "/home/training1/anaconda3/envs/cube/lib/python3.7/site-packages/torch/autograd/grad_mode.py", line 26, in decorate_context
return func(*args, **kwargs)
File "/home/training1/cube_detector/yolov5/val.py", line 211, in run
out = non_max_suppression(out, conf_thres, iou_thres, labels=lb, multi_label=True, agnostic=single_cls)
File "/home/training1/cube_detector/yolov5/utils/general.py", line 565, in non_max_suppression
conf_angle, j_angle = x[i, 5+nc:].max(1, keepdim=True)
RuntimeError: cannot perform reduction function max on tensor with no elements because the operation does not have an identity

Can you let me know the annotation format?

Can you help me fix this please ?