Rahul Autade

Abstract

In this project I recreate the HRNet architecture using PyTorch, ensuring an accurate representation of its structure and capabilities. This foundational implementation will serve as the basis for our subsequent objectives. Once the HRNet backbone is established, it can be enhanced it with various classifiers to make it adaptable for different computer vision tasks such as object detection, image classification, human pose estimation, and segmentation realizing the true power of such a multi-modal model in visual recognition challenges. Simultaneously, I created a fusion model by combining HRNet with ResNet, specifically tailored for image classification tasks. This hybrid architecture aims to leverage the fine-grained feature extraction of HRNet and the stability of ResNet's residual connections to achieve a powerful model capable of capturing intricate details while maintaining robustness.