This repository contains the scripts to develop a image classifier in PyTorch.
Utils provides functions for following list of activites.
This python script provides the neccessary functions for preparing the dataset by doing the following operations
- Builds a PyTorch Dataset using downloaded MNIST dataset
- DataLoader which includes data transformation
- provides function for plotting a
xnumber of images from a batch.
- provides utils function to calculate the accuracy of the model based on the prediction and actuals of target values.
This script helps us create a PyTorch convolution model with following architecture and it has the functions for training and testing the script. Also helps us with plotting the performance curves.
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Conv2d-1 [-1, 32, 26, 26] 320
Conv2d-2 [-1, 64, 24, 24] 18,496
Conv2d-3 [-1, 128, 10, 10] 73,856
Conv2d-4 [-1, 256, 8, 8] 295,168
Linear-5 [-1, 50] 204,850
Linear-6 [-1, 10] 510
================================================================
Total params: 593,200
Trainable params: 593,200
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.00
Forward/backward pass size (MB): 0.67
Params size (MB): 2.26
Estimated Total Size (MB): 2.94
----------------------------------------------------------------
This notebooks high level scripts to build the overall pipeline tofor building the Image classifier and render visualizations of images and model performance.