avijit9 / forces

PyTorch implementation of "What happens if..." Learning to Predict the Effect of Forces in Images, ECCV'16

"What happens if..." Learning to Predict the Effect of Forces in Images

This is the source code for a deep net that predicts the effect of applying a force to an object shown in a static image.

Citation

If you find the code useful in your research, please consider citing:

@inproceedings{mottaghiECCV16,
    Author = {Roozbeh Mottaghi and Mohammad Rastegari and Abhinav Gupta and Ali Farhadi},
    Title = {``What happens if..." Learning to Predict the Effect of Forces in Images},
    Booktitle = {ECCV},
    Year = {2016}
}

Requirements

• Python 3.5+
• Pytorch 0.4.0+
• Tensorboard

Training

Prepare dataset:

./upload_dataset.sh

To train the model, run

python train.py

Check the argument list to set hyperparatmers and paths.

Test

To test, run

python test.py

I am getting accracy of 20.2% with Alexnet as encoder. If you can experiment with Resnet and other variants, please send a pull request.

This implementation is based on original lua implementation which can be found at https://github.com/allenai/forces/