Person Segmentation — Entertainment / Media AI
Built a cloud-based person segmentation system for an media client, enabling automated separation of subjects from backgrounds at scale on AWS. The pipeline processes high volumes of video and image content — handling diverse studio conditions, costume variation, and complex scene composition — delivering clean segmentation masks used in post-production, live broadcast, and interactive media workflows. Below was a test done with my own image.
Tools & Architecture Used: - Python - Opencv - Tensorflow Keras - UNET Architecture - AWS - Optimised model for performance - Docker - FastAPI - Github Actions
Thiyaga Bot