Pytorch-Notebooks

A Collection of Jupyter Notebooks with Deep Learning Models created using Pytorch for Computer Vision (Image Classification) problems trained on GPU.

Following are the models created -

• Simple Logistic Regression on MNIST dataset for Digit Recognition with accuracy of 86.2%

• Two layered Neural Network (1 hidden and 1 output layer) on MNIST dataset, trained on GPU using Google Colab with accuracy of 96.9%

• Convolutional Neural Network with 6 Convolutional layers (followed by ReLU activations), 3 Max Pooling layers and 3 Linear layers for CIFAR10 dataset for object classification trained on GPU using Google Colab with accuracy of 73.6%

• A state of the art Convolutional Neural Network model with Residual blocks, ResNet9 that has 8 Convolutional blocks and 1 Linear block for CIFAR10 dataset for object classification trained on GPU using Google Colab improving the accuracy upto 90%

Datasets Used -

• MNIST dataset for Handwritten digits.

• CIFAR10 dataset for 10 types of Objects.