Problem with shape with running detect.py example

[Dominux]

I cloned repo and ran:

python detect.py --cfg cfg/yolov3-spp.cfg --weights yolov3-spp.weights

And got that:

Namespace(agnostic_nms=False, cfg='cfg/yolov3-spp.cfg', classes=None, conf_thres=0.3, device='', fourcc='mp4v', half=False, img_size=416, iou_thres=0.5, names='data/coco.names', output='output', save_txt=False, source='data/samples', view_img=False, weights='yolov3-spp.weights')
Using CUDA device0 _CudaDeviceProperties(name='GeForce MX230', total_memory=2048MB)

Traceback (most recent call last):
  File "detect.py", line 179, in <module>
    detect()
  File "detect.py", line 28, in detect
    load_darknet_weights(model, weights)
  File "d:\My_scripts\Fashion is my profession\Projects versions\D\yolov3\models.py", line 363, in load_darknet_weights
    conv_w = torch.from_numpy(weights[ptr:ptr + num_w]).view_as(conv_layer.weight)
RuntimeError: shape '[32, 3, 3, 3]' is invalid for input of size 255

I have:

• Windows 10
• Nvidia GeForce MX230 with 2GB VRAM
• Cuda 10.1
• Python 3.7.5
• PyTorch with "Today's version"

Any idea?

[glenn-jocher]

@Dominux that’s odd. Can you try the same command with the pytorch weights instead: yolov3-spp.pt

[Dominux]

python detect.py --cfg cfg/yolov3.cfg --weights yolov3.weights

Namespace(agnostic_nms=False, cfg='cfg/yolov3.cfg', classes=None, conf_thres=0.3, device='', fourcc='mp4v', half=False, img_size=416, iou_thres=0.5, 
names='data/coco.names', output='output', save_txt=False, source='data/samples', view_img=False, weights='yolov3.weights')
Using CUDA device0 _CudaDeviceProperties(name='GeForce MX230', total_memory=2048MB)

Traceback (most recent call last):
  File "detect.py", line 179, in <module>
    detect()
  File "detect.py", line 28, in detect
    load_darknet_weights(model, weights)
  File "D:\My_scripts\Fashion is my profession\Projects versions\D\yolov3\models.py", line 363, in load_darknet_weights
    conv_w = torch.from_numpy(weights[ptr:ptr + num_w]).view_as(conv_layer.weight)
RuntimeError: shape '[32, 3, 3, 3]' is invalid for input of size 255

[glenn-jocher]

@Dominux the .weights are darknet weights. .pt are pytorch weights. You might try re-running with pytorch weights, or running your code in a working envorionment:

• Your environment. If your issue is not reproducible in a GCP Quickstart Guide VM we can not debug it. Ensure you meet the requirements specified in the README: Unix, MacOS, or Windows with Python >= 3.7, PyTorch >= 1.4 etc. You can also use our Google Colab Notebook and our Docker Image to test your code in a working environment.

[FranciscoReveriano]

you might want to try "weights/yolov3-spp.weights"

[Dominux]

you might want to try "weights/yolov3-spp.weights"

It doesn't work

[glenn-jocher]

@Dominux it may just be a windows-related issue then. We routinely validate the code on MacOS and Linux, but our most recent Windows validation was several months ago. It's possible the changes since then may cause a problem on Windows.

Unfortunately we don't have a current Windows machine to test on, nor time to spin up a Windows VM, but if you discover the source of the issue please let us know.

[Dominux]

@glenn-jocher do u know if I can run it from windows subsystem for linux?

[glenn-jocher]

@Dominux I'm not sure, but we do have various working linux environments available you can use and access from windows, many free. We will try to validate on windows at some point soon, but in general we highly recommend you try to develop under linux, even moreso for PyTorch CUDA and GPU accessibility and operability. https://docs.ultralytics.com/yolov5/tutorials/train_custom_data

Reproduce Our Environment

To access an up-to-date working environment (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled), consider a:

• GCP Deep Learning VM with $300 free credit offer: See our GCP Quickstart Guide
• Google Colab Notebook with 12 hours of free GPU time: Google Colab Notebook
• Docker Image from https://hub.docker.com/r/ultralytics/yolov3. See Docker Quickstart Guide

[glenn-jocher]

I'll close this issue for now as the original issue appears to have been resolved, and/or no activity has been seen for some time. Feel free to comment if this is not the case.