chrome-vision

Chrome Vision is a computer vision model that improves the contrastive learning framework MoCo’s accuracy by detecting and cutting images before classification.

Prerequisites

Setup conda environment

Create environment

conda env create -f environment.yml
# conda env remove -n cv

List environment

conda env list

Activate enviroment

conda activate cv
# conda deactivate

Install PyTorch

If using CPU

conda install cpuonly -c pytorch

If using GPU with CUDA support

conda install pytorch-cuda=11.6 -c pytorch -c nvidia

Usage

Run chrome vision inside the train.ipynb jupyter notebook