Skip to content

Latest commit

 

History

History
30 lines (18 loc) · 1.28 KB

README.md

File metadata and controls

30 lines (18 loc) · 1.28 KB
Raw

cam-pos-system

cam-pos-system is a Rust library with an example app that reads a video stream and detects moving objects within a coordinate system using computer vision algorithms.

moving train

Activity

Algorithms

The cam-pos-system library uses the following computer vision algorithms for object detection:

  • HOG detector: This algorithm detects the presence of objects in an image by looking for their Histogram of Oriented Gradients (HOG) features.
  • MosseTracker: This algorithm tracks a moving object in a video stream using adaptive correlation filters to estimate the object's position and scale. The library combines these algorithms to detect moving objects in a video stream and estimate their positions within a coordinate system.

run wasm

wasm-pack build --no-default-features --target web
python3 -m http.server

training

In the folder res/training are images. For each image exists a txt file containing labels in format "name x y" The label names can be found in res/labels.txt.

annotator tool

You can create the training data with this simple tool on Hog-Detector.