Image classification on MNIST and CIFAR10
Image Classification on MNIST handwritten digit recognition and CIFAR-10
Authors
Prasheel Renkuntla
Description
This project demonstrates image classification applied on MNIST and CIFAR10 dataset. For both datasets, the Neural Network architecture is build on top of different existing architectures like LeNet5, ResNet.
Dependencies
- Python 3.7
- TensorFlow v2.0.0
- Matplotlib
- sklearn
- scipy
- time
- sys
- os
- seaborn
- If on PC, tensorflow-gpu would be faster.
Run
On Google Colab (with GPU, Runtime -> Change RunTimeType -> GPU) Colab files can be directly loaded
To run Classification task on MNIST Data- Without GPU, Average Runtime is 10 mins approx.
For ready code-
python3.7 ready_code_mnist.py
For Architecture like LeNet-
python3.7 my_way_mnist_LeNet5.py
For my Architecture-
python3.7 mnist_classification_my_arch.py
To run Classification task on CIFAR-10 Data- (GPU Preferred!) (similar names .ipynb also provided)
For ResNet 20-
python3.7 cifar10_resnet20_25.py
For ResNet 32-
python3.7 cifar10_resnet32_200.py
For ResNet 56-
python3.7 cifar10_resnet56_200.py