sroy729/SoC-CNN-lytical

Assignments completed under SoC

SoC-CNN-lytical

The following are the brief details of the models implemented in this repository

• Assignment 1 has implementation of image classification for MNIST dataset from scratch. It uses the following python Libraries:

  1. numpy arrays are used for simpler code and faster computations
  2. matplotlib.pyplot is used for plotting images and visualizing data and results
  3. sklearn.model_selection.train_test_split is used for splitting the dataset into train and test dataset

  The final accuracy of the trained model has an accuracy of 95.067 %.
  The above libraries are used in all the model implementations.

• Assignment 2 has implementation of image classification for MNIST dataset. It uses torch library for easier and faster implementation of Neural networks.

  The final accuracy of the trained model has an accuracy of 97.723 %.

• Assignment 3 has implementation of image classification for CIFAR-10 dataset using Convolutional Neural Networks. It uses torch for Dataloader class as well as for torch.nn.Module class for neural network. torchvision is used for transforming images.
  The final accuracy of the trained model is 66.242 %.

• Assignment 4 is about image segmentation on Caravana dataset. It implements U-Net for the same. It also used torch and torchvision libraries.
  The final accuracy of the trained model has an accuracy of 89.714 %.