IMaterialist classifier

Description

Prerequisites

You will need the following things properly installed on your computer.

• Docker

Installation

• git clone https://github.com/smorzhov/imaterialist.git

Running

Remember that Docker container has the Python version 3.5.2!

1. If you are planning to use nvidia-docker, you need to build nvidia-docker image first. Otherwise, you can skip this step

   nvidia-docker build -t sm_keras_tf_py3:gpu .

   Run container

   nvidia-docker run -v $PWD/src:/imaterialist -dt --name imc sm_keras_tf_py3:gpu /bin/bash

2. Download test, train and validation data

   nvidia-docker exec imc python3 download.py data/test.json data/test
   nvidia-docker exec imc python3 download.py data/train.json data/train
   nvidia-docker exec imc python3 download.py data/validation.json data/validation
   

3. Training

   By default, only the 0th GPU is visible for the docker container. You can change this by passing --env option to exec. For example:

   nvidia-docker exec --env CUDA_VISIBLE_DEVICES='0,1,2' imc python3 train.py

   This will start training on 0th, 1st and 2nd GPUs.