Object recognition using deep ConvNet and Caltch-256 dataset
This project is my first foray in learning and implementing deep convolutional neural network using Tensorflow.
Need python 2.7
Data -> Contains image data (both test and train)
vcl_setup.sh - Run it to setup tensorflow with GPU support and all important python packages in Ubuntu machine
read_data2.py -> Reads images for given folder and range and augments them using ritation and gaussian blur.
read_data_multiple_category.py -> Wrapper for reading images from multiple folders
cnn_*.py -> Actual CNN network. Initial code is to read the images and create label matrix. Will separate this module out later. cnn_6.py has a provision of stopping the code when the training accuracy flatlines after certein number of iterations.Please use cnn_5.py or cnn_6.py.
To test, just run cnn_6.py.
- All the references are mentioned in projectPaper.