Arijit-datascience / pytorch_cifar10

CIFAR10 image recognition using ResNet architecture, Gradcam images

Code base for Object detection on CIFAR10 using Pytorch

This repo is the internal engine for our Object detection projects on CIFAR10 using Pytorch. Lets take a look at the components:

model

Here you can find code structure for all the models that we have used. Following is the list, for now:

• Resnet18
• custom_resnet

utils

Home for code related to processing and augmentation of images. Current residents:

• gradcam.py

  Gradcam code to visually understand parts of the image that our model is focusing on.

  Reference: Kazuto Nakashima

• utils.py

  Contains following processes:

  • Mean and Standard Deviation calculation
  • Image Transformation
  • Downloading the Dataset
  • Dataloaders
  • Display sample images from the dataset
  • Plot loss and accuracy graphs
  • Seeding
  • Display images that were misclassified

• lr_finder.py

  Calculate the best max_lr parameter for OneCycleLR.

main.py

You can find all the major orchestration code here. Currently available:

• train
• test
• main

This repo is constantly being updated based on all the new and intersting Computer Vision tasks we do. Do visit again!

Contributors

Abhiram Gurijala
Arijit Ganguly
Rohin Sequeira