EgoCentricGestureNet

A repo with model and links to dataset from Egocentric Gesture Recognition for Head-Mounted AR devices (ISMAR 2018 Adjunct)

Steps to test the model

• Make sure you all the dependencies below installed
• Run download_dataset.sh script, this will download the dataset(EgoCentricGestures.tar.gz) to current directory
• Move this file to an appropriate location and untar the file. This will generates two directories and README.txt files explaining the structure of the dataset and directories
• run the model with 'python ego_gesture_net_test.py --test_dir=<full_path_to_test_directory>'. This should output the actual gesture id and the recognised id by the network.

Dependencies

• wget
• python - 3.5
• torch - 0.3.1
• torchvision - 0.2.1
• Pillow - 5.2.0
• cuda - 8.0.61
• nvidia driver - 384.59

Links

• Dataset
• Preprint

Citation

• Bibtex Entry

@inproceedings{tejo2018ismar, author = {Tejo Chalasani and Jan Ondrej and Aljosa Smolic}, title = {Egocentric Gesture Recognition for Head-Mounted AR devices}, booktitle = {2018 {IEEE} International Symposium on Mixed and Augmented Reality, {ISMAR} 2018 Adjunct}, year = {2018} }