liaorongfan / UBKProj

UBK Project

Set up experiments on deep learning models with UBK data

Installation

pip install torch
pip install torchsampler

Data preparation

Data should be processed first and put aside in dataset directory. The raw data and processed data are made available on request only.

Usage

We provide MLP, AlexNet, Vgg16, Resnet18, Transformer, and Swin-transformer models for evaluation, you can conduct experiments and select models by --config-file argument shown as below. The corresponding config files are listed in config directory.

Quick start

# run with the demo data in git repo 
python run.py -c config/alexnet_demo.yaml

Training

# cd path/to/UBKProj
python run.py --config-file config/vgg16_1d.yaml 

Resume training

python run.py --config-file config/vgg16_1d.yaml TRAIN.RESUME /path/to/weight

Test

python run.py --config-file config/vgg16_1d.yaml --test-only TEST.WEIGHT /path/to/weight