hhhhhli / garofv

GarOFV

This code is for our paper "Continuous Perception Garment ClassificationBased on Optical Flow Variation".

install

git clone https://github.com/hhhhhli/GarOFV.git
cd GarOFV
pip install -r requirements.txt  # install

preparation

The dataset can be downloaded from the link below:

https://github.com/LiDuanAtGlasgow/GarNet

The image-based classification model can be acquired according to GarNet.Which can be replaced by other image_based classification methods.

If you need to change the dataset and classification model paths, Modify the video_classify_module.py file.

354    file_path='./Database/'
355    model_path='./Models/'

Run the code

You can use command line arguments

    --bandwidth     
    --use_flow
    --use_early_stop

to modify bandwidth, whether to use our optical-flow based method and whether to use early-stop model.

For example, to run in continuous perception model with our method, and set the bandwidth is 95, can use the following bash command:

python -B video_classify_module.py --bandwidth 95 --use_flow True --use_early_stop False

To run in early-stop model with our method, and set the bandwidth is 55, can use the following bash command:

python -B video_classify_module.py --bandwidth 55 --use_flow True --use_early_stop True

Can run the run.sh file to obtain the run result in different modes and bandwidth

bash run.sh

The results are saved in the Results directory, and run the line_chart.py file can get the visualization results.

python line_chart.py

References and acknowledgement

Our code refers to Li Duan's GarNet algorithm.

https://github.com/LiDuanAtGlasgow/GarNet