Embedded Files
StreamSense AI was my Capstone project for my Masters Program, completed in Spring 2026. It is a Machine-Centric Video Streaming Pipeline for Ai Inferences, and uses technologies like FFMPEG, YOLO-World, and Optical Flow
StreamSense AI was my Capstone project for my Masters Program, completed in Spring 2026. It is a Machine-Centric Video Streaming Pipeline for Ai Inferences, and uses technologies like FFMPEG, YOLO-World, and Optical Flow
Overview
Overview
● I was the Team Lead of 4. This video streaming pipeline maintains
● I optimized the h.264 encoding process to intermittently adjust compression strength based on Object Detection confidence reports. My method never drops frames and can reduce latency by 30%.
● My full pipeline saves 80% Bandwidth Consumption with less than 7% average confidence loss.
● My team met all milestones and earned high praise at the UCI ICS Expo because of my management and effective poster design.
Showcase
Showcase
The video highlights our Capstone Poster and project architecture, which features a feedback loop that continuously adjusts video stream quality to reach a specified target confidence level. With our filtering and dynamic encoding process, we can use up to 82% less bandwidth consumption than the original video, while maintaining similar accuracy.
Two features to point out. First, the pipeline is designed to support any object detection model that reports back confidence scores and bounding boxes. While our team used Yolo-World for our examples, any team can upload a custom model to better fit their custom use case. Second, unlike other machine-centric video streaming pipelines that can accidentally filter out objects of interest, our filtering system is more transparent and immediately reacts to motion to minimize false negatives and be ready for critical missions.
Our full report can be found here. Please read it if you are interested in how we overcame the hurdles, our results in detail, and how we will apply our findings in the future.
Page updated
Report abuse