After experimenting with various object detection integrations like Frigate, Deepstack, Viseron, and Doods, I found that none of them truly satisfied my needs for accurate night person detection. That’s when I discovered Yolo v7, which not only offers superior accuracy but also seamlessly integrates with Nvidia GPU via CUDA. I’m now on a quest to create a seamless integration within Home Assistant using Yolo v7.
Here’s what I’m aiming to achieve: I want to take the output from detect.py using an IP camera’s RTSP address and convert it into sensors that indicate the detection of specific objects (like persons and cars) with an ON/OFF or 1/0 state. Additionally, I’d love to capture images of detected objects and send them via Telegram.
I’m seeking advice on two main fronts:
- Proper parsing of
detect.pyresults to make them understandable within Home Assistant, specifically focusing on detecting only persons and cars. - Guidance on capturing a detected object’s image and utilizing it within Home Assistant.
If anyone has experience with similar integrations or can provide insight into these challenges, I’d be incredibly grateful! Let’s collaborate to make this integration a reality.