- python 3
- PyTorch
Please note the versions to be installed
torch==0.1.11._5
torchvision==0.1.9
- Download the data into train_module/data
- Fowllow the train_module/preprocess/Readme.md
- Follow the train_module/Readme.md to run train_model.py file.
The models are present in a drive link mentioned below
https://drive.google.com/drive/folders/1xH3zcEEbIePlnW8G029Lqe4nXeWp5AOa?usp=sharing
Download them and paste them in test_module/models
- Download the models from the drive link given above.
- Install the requirements by running pip install -r requirements.txt
- Inside the test_module run the line python3 FE.py for frontend based model
- Run the following command for graphical based model
python run_model.py --image img/CLEVR_val_000013.png --question "Does the small sphere have the same color as the cube left of the gray cube?"
For frontend model please use sudo apt-get install python-qt4 to install pyqt-4 and sudo apt install python3-tk to install tkinter
- Follow the training section to train the model
- Then copy the trained models instead of downloading in the above testing step.
- Follow the rest of the steps mentioned in the testing section
- Stackoverflow
- Clevr-iep[https://github.com/facebookresearch/clevr-iep]
- Clevr-iep[https://arxiv.org/abs/1705.03633]