/Human-Segmentation-using-U-NET

Segmentation using Deep Learning

Primary LanguageJupyter Notebook

Human Segmentation with U-Net, a Deep Learning Algorithm based on CNN

logo

Human Segmentation Using U-NET

Ayush Mohapatra · Divya Swaroop Dash · Shreya Tiwari · Mokshada Mohapatra · Surbhi Negi


Overview of the Project

Segmentation refers to the process of dividing an image or video into distinct sections, with each piece representing a specific object, region, or feature. This technique is crucial for object recognition and analysis. By isolating objects from the background, the computer can more accurately identify and analyze them.

Using the concept of segmentation, we can perform human segmentation, which separates humans (bodies and faces) from the background. This facilitates tasks such as object detection, object removal or addition.

This project aims to automatically segment humans (bodies and faces) from images or videos. This segmentation can be used for object detection. The proposed model is built on UNET architecture and had an accracy of 90%

Dataset

The dataset we used in our project: Human Segmentation Dataset - Supervise.ly

Demo

demo