Skip to content

aparwal/Tensorflow_Image_Classifier

Repository files navigation

Tensorflow_Image_Classifier

Image Classification using CNN in TensorFlow

A simple Convolution Neural Network to classify images into classes built from scratch in TensorFlow.

Usage

To test the classifier, run

python test.py

To train the classifier, run

python train.py

Note

Currently trained to classify road bikes and mountain bikes using very few images(~100 per class)
It's not very generalizable right now, but works for the given data
Can be trained on any large custom dataset for better performance
Put the custom dataset in data/train directory with image label as subdirectories

Dependencies

  • Python 3.6
  • Numpy 1.x
  • OpenCV 2.x or 3.x
  • TensorFlow 1.8.0

Description

  • data_preprocess.py: Contains helper functions to read and process images
  • model.py: Contains helper functions to create a tensorflow network
  • results/: Constains saved output of test.py
  • models/: Contains the saved model from train.py
  • tboard/: Contains tensorboard logs from taining and testing

Tensorboard graph

Graph

About

Image Classification using CNN in TensorFlow

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages