This repository contains code for objection recognition in images using Convolutional Neural Networks. It is developed as part of the competition CIFAR-10 hosted by Kaggle. Check out this link for more details : https://www.kaggle.com/c/cifar-10/ and Data
The output folder contains the results of running the network. The best accuracy of 0.79960 was achieved using Convolutional Networks with Rectified Linear Unit. This net had 5 layers, No Zero Padding and Droput of 0.25 between layers and 0.5 at the Final Softmax Layer.
This code is written in python. To use it you will need:
- Python 2.7
- Theano 0.7 (Back end)
- Keras
- scikit-learn (Machine Learning Library in Python)
- A recent version of NumPy and matlplotlib
- imutils (Basic Image Processing functions)