yqyao / yolov3_pytorch Goto Github PK

hello, I meet this problem, I have executed make.sh. And There is a gpu_nms.pyx in the utils folder. This is the tracklog:
`
$ python train.py --input_wh 1024 1024 -b 64 --subdivisions 4 -d VOC --basenet ./weights/convert_darknet53.pth
Traceback (most recent call last):
File "train.py", line 15, in
from model.yolo import Yolov3
File "/home/gmy/lqb/YOLOv3_Pytorch/model/yolo.py", line 11, in
from utils.box_utils import permute_sigmoid, decode
File "/home/gmy/lqb/YOLOv3_Pytorch/utils/box_utils.py", line 11, in
from utils.nms_wrapper import nms
File "/home/gmy/lqb/YOLOv3_Pytorch/utils/nms_wrapper.py", line 9, in
from .nms.gpu_nms import gpu_nms
ModuleNotFoundError: No module named 'utils.nms.gpu_nms'

`

Dear Author,
Excellent source code you provide.
I use your code and test the yolov3.coco pre-trained model perfectly.
However, I use the same procedure on yolov3.openimages pre-trained model, the outputs are very poor: it predict so many classes and so many boxes, but none of them is correct.
And I didn't find the bug yet.

Do you know any difference between the yolov3.coco and the yolov3.openimage?
I just found that their output layers and the channel nums of one convolutional layer are different, i.e. 80vs601, 255vs1818
The pre-trained weights can be load, the networked can run, but I just didn't found the bug that causes the very poor performance.

Hello,thanks very much for your sharing.
I downloaded your code and wanted to do my experience.But I didn't find the wieght ('convert_yolov3_voc.pth'),also didn't find any information how I can get this model.
Can you give me some help?Thank you!!!

I've been trying to create 'convert_yolov3_coco.pth' by following the README.
But the following error occurs.
Looking into the directory, utils.nms, there is cpu_nms.pyx instead cpu_nms.py.
How can I import DEFs from .pyx file??
Please teach me how to address this problem.

@user:~/git/YOLOv3_Pytorch$ python -m convert_darknet.py --version coco --weights ./weights/yolov3.weights --save_name ./weights/convert_yolov3_coco.pth
Traceback (most recent call last):
  File "/home/nel/anaconda3/envs/CenterNet_ONNX/lib/python3.6/runpy.py", line 183, in _run_module_as_main
    mod_name, mod_spec, code = _get_module_details(mod_name, _Error)
  File "/home/nel/anaconda3/envs/CenterNet_ONNX/lib/python3.6/runpy.py", line 109, in _get_module_details
    __import__(pkg_name)
  File "/home/nel/git/YOLOv3_Pytorch/convert_darknet.py", line 10, in <module>
    from model.yolo import Yolov3
  File "/home/nel/git/YOLOv3_Pytorch/model/yolo.py", line 11, in <module>
    from utils.box_utils import permute_sigmoid, decode
  File "/home/nel/git/YOLOv3_Pytorch/utils/box_utils.py", line 11, in <module>
    from utils.nms_wrapper import nms
  File "/home/nel/git/YOLOv3_Pytorch/utils/nms_wrapper.py", line 8, in <module>
    from .nms.cpu_nms import cpu_nms, cpu_soft_nms
ModuleNotFoundError: No module named 'utils.nms.cpu_nms'

I am trying to convert yolo-tiny.weights to yolo-tiny.pth but I'm getting the following error can anyone please help

8858734
Traceback (most recent call last):
File "convert_darknet.py", line 243, in
load_weights(weightfile, yolov3, version)
File "convert_darknet.py", line 102, in load_weights
load_weights_yolov3(weightfile, yolov3)
File "convert_darknet.py", line 137, in load_weights_yolov3
ptr = copy_weights(bn, conv, ptr, weights)
File "convert_darknet.py", line 62, in copy_weights
conv_weights = conv_weights.view_as(conv.weight.data)
RuntimeError: shape '[512, 256, 3, 3]' is invalid for input of size 391694