Harshith Jai Surya Ganji - 230041010 (team leader) Mullapudi Namaswi - 230041023 Korubilli Vasihnavi - 230041016
This project is a Streamlit application that allows users to replace the background of a video with a custom background image using a pre-trained DeepLabV3 model for segmentation.
The application uses a pre-trained DeepLabV3 model from PyTorch to segment the foreground (person) from the background in a video. The segmented foreground is then combined with a user-provided background image to generate a new video with the replaced background.
- Upload a video and a background image.
- Replace the video background with the uploaded background image.
- Download and view the processed video directly in the browser.
-streamlit -opencv (python) -torch -torchvision -numpy -pillow -tempfile
- a pre-trained model (deeplabv3) is used for segmentation
- a pipeline is built to preprocess frames of video
- helper function takes input and applies masking on them, separating backgrounds and foregrounds
- a second function is used to combine required background and foreground to produce output
Streamlit for providing the framework for this application. PyTorch for the DeepLabV3 model and other utilities. COCO dataset for providing the segmentation classes.
Pipeline and model implementation - Namaswi and Vaishnavi Helper functions - Harshith