Skip to content
A simple framework to automate image classification
Python Dockerfile Makefile Shell
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
docker
.gitignore
LICENSE
Makefile
README.md
autokeras.sh
download_cifar10_images.py
inception.sh
train_autokeras_model.py
train_inception_model.py

README.md

auto-classify-images

A simple framework to automate image classification. Simply create the ./image/train and ./image/test directories with labeled subdirectories, and then fill each subdirectory with representative images. Then run one of the two shell scripts below to begin training (or both).

Loading cifar10 example data

$ mkdir -p images/train # Optional

$ mkdir -p images/test # Optional

$ python download_cifar10_images.py # Download cifar10 files

Prerequisites

The container for this project is based on the nvidia/cuda docker implementation. Thus, prerequisites for running this container are equivalent to those at https://hub.docker.com/r/nvidia/cuda/ .

Note: The container should still run with only a CPU (i.e. no GPU), albeit slowly.

Train using Inceptionv3 transfer learning

$ ./inception.sh

Train using AutoKeras Neural Architecture Search

$ ./autokeras.sh

You can’t perform that action at this time.