Image Classification using CNN in TensorFlow
A simple Convolution Neural Network to classify images into classes built from scratch in TensorFlow.
To test the classifier, run
python test.py
To train the classifier, run
python train.py
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
- Python 3.6
- Numpy 1.x
- OpenCV 2.x or 3.x
- TensorFlow 1.8.0
data_preprocess.py: Contains helper functions to read and process imagesmodel.py: Contains helper functions to create a tensorflow networkresults/: Constains saved output of test.pymodels/: Contains the saved model from train.pytboard/: Contains tensorboard logs from taining and testing
