Running a light-weight onnx model in react to detect boxes on an image

[ethanmaurer]

Hello, I am currently working on a react project with the goal of analyzing an image and returning an array of bounding boxes which would include the x, y, with, and height of each box. I have a solution currently using onnxrunntime-web and the 'segment-anything' model (SAM). The main issue with this model is it's very slow to run and takes a good 8-10 seconds per image and I would like to pass 10+ images in. Does anyone know of a lighter-weight model I could use? It does not have to be very advanced as all the images being passed in only include boxes of various sizes so it just needs to detect the edges of all of the boxes. It also needs to run on the browser exclusively. Any ideas?