Skip to content
/ EVA4 Public

Solution to the assignments in the EVA4 course

Notifications You must be signed in to change notification settings

rvk007/EVA4

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

43 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TSAI - Extensive Vision AI 4

This repository contains the solutions to the assignments of the EVA4 course conducted by The School of AI.

Contents

Architectural Basics (MNIST Classification)

Reaching 99.4% accuracy on the MNIST test dataset with a model having less than 20,000 parameters and has been trained for less than 20 epochs. To see the code go here.

Coding Drill Down

Reaching 99.4% accuracy on the MNIST test dataset with a model having less than 10,000 parameters and has been trained for less than 15 epochs. To see the code go here.

Regularization

Applying L1 and L2 regularization on the final model trained in Session 5. To see the code go here.

Advanced Convolutions

Reaching a test accuracy of 80% on CIFAR-10 dataset using advanced convolutions. To see the code go here.

Receptive Fields and Network Architecture

Reaching a test accuracy of 85% on CIFAR-10 dataset with ResNet18 model. To see the code go here.

Data Augmentation and Grad Cam

Reaching a test accuracy of 87% on CIFAR-10 dataset with ResNet18 model using Grad Cam and various data augmentation techniques. To see the code go here.

LR Finder and Reduce LR on Plateau

Reaching a test accuracy of 88% on CIFAR-10 dataset with ResNet18 model using LR Finder and Reduce LR on Plateau. To see the code go here.

Super Convergence

Reaching a test accuracy of 90% on CIFAR-10 dataset custom ResNet model using One Cycle Policy for Learning Rate. To see the code go here.

Tiny-ImageNet and YOLO v2 Anchor Boxes

Reaching a test accuracy of 50% on Tiny-ImageNet dataset with ResNet18 model and finding the anchor boxes for YOLO v2 using K-Means Clustering algorithm. To see the code go here.

Object Detection with YOLO v3

Using transfer learning to detect a custom object using YOLO v3. To see the code go here.

Segmentation and Depth Estimation Dataset Creation

Creating a dataset with 400,000 images for image segmentation and depth estimation. To see the code go here.

Image Segmentation and Depth Estimation

Creating a model which can perform image segmentation and depth estimation on a custom dataset. To see the code go here.