requirements.txt

### Setup environment timeSformer
Step 1:
conda create -n timesformer python=3.7 -y
source activate timesformer

Step 2:
conda install pytorch=1.8.1 torchvision=0.9.1 cudatoolkit=10.2 -c pytorch

Step 3:
pip install 'git+https://github.com/facebookresearch/fvcore'

Step 4:
pip install simplejson
pip install einops
pip install timm

Step 5:
conda install av -c conda-forge

Step 6:
pip install psutil
pip install scikit-learn
pip install opencv-python-headless
pip install tensorboard

Step 7:
pip install pyyaml
pip install munch
pip install matplotlib

Step 8:
select correct interpreter (the conda timesformer environment) in vscode

Step 9:
clone repository TimeSformer https://github.com/facebookresearch/TimeSformer.git and add folders "timesformer" and "tools" to project
replace train_net, test_net, vit.py, meters.py, checkpoint.py with the ones in this repository

Step 10:
set path to TimeSformer_divST_8x32_224_k600.pyth
set path to datasets

Step 11:
run timesformer_call.py


### Setup environment DenseNet
conda create -n densenet python=3.7 -y
conda activate densenet

pip install tensorflow
pip install matplotlib
pip install sklearn