search

ChristianMarzahl / ObjectDetection

Some experiments with object detection in PyTorch
136 stars 43 forks source link

Exception: It's not possible to apply those transforms to your dataset #21

Closed madrez closed 4 years ago

madrez commented 4 years ago

I receive the following error when trying to build my dataset: Exception: It's not possible to apply those transforms to your dataset: Expected object of scalar type Float but got scalar type Double for argument #3 'mat2' in call to _th_addmm_out

This problem has never occurred until a few hours ago. And I am using the exact code, directory and files as before. Do you think it has occurred because of an update?

madrez commented 4 years ago

The issue was because of the new version of PyTorch and Fastai incompatibility. Temporary fix is by: !pip install "torch==1.4" "torchvision==0.5.0

ChristianMarzahl commented 4 years ago

Thank you for this information. If Fastai is compatible with the new PyTorch version I will update my object detection repo.

madrez commented 4 years ago

Thanks @ChristianMarzahl thanks for you reply. I am using your repo for a research project. By the way have you seen YOLOv4? Are you considering adding it?

ChristianMarzahl commented 4 years ago

I have seen YOLOv4 but at the moment absolutely no time add something. What type of research are you doing?

If you also need to label and annotate images have a look at https://github.com/ChristianMarzahl/Exact Its interacts nicely with this repo.

With kind regards, Christian

madrez commented 4 years ago

YOLOv4 sounds very promising. Can't wait to see you Fastaiify it :) Medical imaging. I am developing a model to detect endoscopy polyps. Got my dataset cleaned up and trying to tune it as much as I can. Thanks! I'll have a look at that. Currently using "RectLabel" mac app for annotation.