heyvi 0.2.3

Project Vi

"Hey Vi!"

HEYVI: Visym Analytics for Visual AI
docs: https://developer.visym.com/heyvi

HEYVI is a python package for visual AI that provides systems and trained models for activity detection and object tracking in videos.

HEYVI provides:

• Real time activity detection of the MEVA activity classes
• Real time visual object tracking in long duration videos
• Live streaming of annotated videos to youtube live
• Visual AI from RTSP cameras

Requirements

python 3.*
ffmpeg (required for videos)
vipy, torch, pytorch_lightning (for training)

Installation

pip install heyvi

Quickstart

v = heyvi.sensor.rtsp().framerate(5)
T = heyvi.system.Tracker()
with heyvi.system.YoutubeLive(fps=5, encoder='480p') as s:
     T(v, frame_callback=lambda im: s(im.pixelize().annotate()))

Create a default RTSP camera and GPU enabled object tracker, then stream the privacy preserving annotated video (e.g. pixelated bounding boxes with captions) to a YouTube live stream.