React video player and konva
Motivation
I work with very talented data scientists and engineers. A lot of the models they are building are used to identify objects and actions in a video. Their models produce raw results in a CSV, JSON format. It is difficult to validate their results just from looking at the raw data. We have tried to feed the data into 3rd party tools to visualize the data in graphs, and there are a lot of python projects that can draw bounding boxes on images or videos .. but there doesn't seem to be a good solution to visualize bounding boxes on a video on a website.
Interactive video player that overlays live computer-vision bounding boxes on playback using React, Konva, and Material-UI. Lets reviewers see exactly what an ML model sees, frame by frame — essential for tuning detection thresholds.