SHREC2024 - Hand Motion Recognition

Description

This repository contains the submission, i.e. the code and models, by

• Martin Hanik,
• Esfandiar Navayazdani, and
• Christoph von Tycowicz

to the Recognition Of Hand Motions Molding Clay track of the 3D Shape Retrieval Challenge (SHREC) 2024. The task is to classify highly similar hand motion sequences from a professional potter into one out of seven classes. The sequences are of variable length and comprise 3D coordinate data of landmarks on both potters hands.

Running the code

You can start a live demo of the prediction on binder:

To run the test script on all test sequences execute the following

python test.py --path='./data/Test-set/**/*.txt'

The code that generates an ensemble based on random training/validation splits can be found in train.py.

Data

The data is provided by the organizers of the SHREC 2024 challenge and can be downloaded here. To set up the data, you can run the start script (this will happen automatically when starting the binder instance; see above)

./start

Afterward, the data is provided in the data folder. The data is split into a training and a test set. The training set contains 50 sequences, the test set 12 sequences. Each sequence is stored in a separate text file. The text files contain the 3D coordinates of the landmarks of both hands.