Convolution Neural Network Demonstration and Visualization
- This CNN was designed to classify images as either Dog or Cat.
- Trained using a total 8000 samples of dogs (4000) and cats (4000). Tested with 2000 samples of cats and dogs (1000 each).
- You may use your own samples for training and testing.
- Uses Keras with either tensorflow or theanos as a backend. Currently only works with CPU or Nvidia CUDA.
- Acheived about 85% accuracy with 2 convolution layers (with 32 filters of size 3x3) and 2 pooling layers (with a pool size of 2x2)
