michaeldeyzel / skeleton-dml

Skeleton-DML: Deep Metric Learning for Skeleton-Based One-Shot Action Recognition

Skeleton-DML

This repository contains the source code to reproduce the results from the Skeleton-DML paper. A pre-print can be found on arxiv.

Video Abstract

Requirements

• Skeleton-DML is based on the pytorch-metric-learning library

Precalculated Representations

We provide precalculated representations for all conducted experiment splits of the Skeleton-DML representation:

• NTU RGB+D 120 One-Shot

Quick Start

git clone https://github.com/raphaelmemmesheimer/skeleton-dml
cd skeleton-dml
pip install -r requirements.txt
export DATASET_FOLDER="$(pwd)/data"
mkdir -p data/ntu/
wget https://agas.uni-koblenz.de/datasets/skeleton-dml/skeleton-dml-ntu_120_one_shot.zip
unzip skeleton-dml-ntu_120_one_shot.zip -d $DATASET_FOLDER/ntu/ntu_reindex
python train.py dataset=ntu_reindex

when returning you have to set the dataset folder again:

export DATASET_FOLDER="$(pwd)/data"
python train.py dataset=ntu_reindex

Note, the following commands require an environment variable $DATASET_FOLDER to be existing.

Training for the NTU 120 one-shot action recognition experiments can be executed like:

python train.py dataset=ntu_reindex

During development, we suggest using the classes A002, A008, A014, A020, A026, A032, A038, A044, A050, A056, A062, A068, A074, A080, A086, A092, A098, A104, A110, A116 as validation classes.