Need to understand how to import weights
abhi-vellala opened this issue · comments
Hello,
Firstly, I'd like to congratulate you for bringing this amazing work. Class agnostic object detection is much needed currently in the industry and this would be a great way to solve the problem.
I wanted to test your model on some custom data. However, I cannot import pre-trained weights from the link you have provided. I can see the zip file but I couldn't find a way to import them. I'm using OpenCV to import weights. It is asking me to have a config file as well as .weights file.
Could you please help me which library to use to import weights when I'm working on a jupyter notebook?
Thank you,
Hi @abhi-vellala,
Thank you for your interest in our work. You can use the script inference/main.py to test our MDef-DETR model on your custom dataset. Please follow the steps below,
- Clone the repository and set the PYTHONPATH environment variable
git clone https://github.com/mmaaz60/mvits_for_class_agnostic_od.git
cd mvits_for_class_agnostic_od
export PYTHONPATH="./:$PYTHONPATH"
- Download the pre-trained weights pretrained_models/MDef_DETR_r101_epoch20.pth
- Run the script as
python inference/main.py -m mdef_detr -i <path to directory containing custom dataset images> -c <path to the downloaded MDef-DETR weights file> -tq 'all objects'
The script will save the class-agnostic detections in .pkl
format at the same path containing your custom dataset images directory. The .pkl
file contains a dictionary with keys
as image names
and values
as the tuple of predicted boxes & scores (i.e. ([boxes], [scores])
). You can also save the predictions in .txt
file using SaveTxtFormat
class in main.py
from save_predictions.py.
Additionally, you can run python inference/main.py -h
to see the complete usage of the script. Moreover, you can try using different general or targeted queries using the flag -tq
. For example, -tq 'all visible entities and objects'
, -tq 'all salient objects'
, etc.
I hope it will help. Let me know if you have any questions.
Hi,
Feel free to reopen if you have any confusion.