Indian-Traffic-Segmentation

To perform Semantic(Image) Segmentation of Indian Traffic Images.

Description:

Semantic segmentation, or image segmentation, is the task of clustering parts of an image together which belong to the same object class. It is a form of pixel-level prediction because each pixel in an image is classified according to a category. This project involves in segmenting the Indian Traffic Images obtained from the video which contains the traffic sites observed in various cities of India. This is implemented using U-Net architecture based Convolutional Neural Network.

Ref for U-Net: https://arxiv.org/abs/1505.04597

Dataset: http://idd.insaan.iiit.ac.in/

Sample Image

Results:

IOU score for Train and Val Data are 0.4715 and 0.4199.

Sample Outputs

Blog link: https://harishkumar-69065.medium.com/indian-traffic-semantic-segmentation-d4a0062c81fc

Contains one file and one folder:

• Segmentation.ipynb - Segmentation using U-Net.
• Images Folder.