Real-time wild boar detection based on yolov5 with video preview

wildboar

This is an open-source project. Hope you like the project and feel free to contribute to it!

4 Likes

Nice! Although when I first read the title I thought "wild boar" was some software api or product. But in fact it is a wild boar. :rofl: :rofl: :rofl:

2 Likes

hahaha, it is a wild boar :boar:

1 Like

@huangyouyou ,
Nice project!
Do you have also a link to it (github repo, or something else...)?
Bart

Thank you for your appreciation!! Feel free to contribute to it. :smiley:

Seeed-Studio/node-red-contrib-ml: Get started with AI vision at the edge with no coding experience at all! (github.com)

2 Likes

Very nice project, well done.

there's never a wild boar around when you want to test in UK !

1 Like

We have badgers and foxes. To be honest, I would love to find out what is digging the big hole in our garden!

But given our recent battles with the local council about road changes, I'd be most interested in having a vehicle counter. :slight_smile:

you don't think data or evidence are going to make the council change their minds about anything do you ?!? - Stick to counting badgers.

1 Like

Actually, if I could demonstrate badgers in the road, I'd probably get a better response :rofl:

2 Likes

I would be lucky if I could use AI to detect parking enforcement officers :wink:

2 Likes

...a set for training the model

Watch out, here they come

3 Likes

Nice project, but when I read "Nvidia" and "Jetson" I read:
Expensive in runtime because eats much energy over time.
I use Yolo on pictures my cams upload ton an FTP and push it to Telegram by bot then.
(My solution for the poor people :slight_smile: )

Interesting. What device do you run the classifier on?

I’m running my Jetson Nano in 5W mode, using YOLOv4. Is not that much energy consumption I think, not much more than a LED lamp

1 Like

wow, 5W is less than I expected. - About how many cams are we talking ? (Jetson can Handle) and the how high is the system workload ?
I upload just singe images only on movement recognition to a V-Server which costs about $7 per month. Then I use Yolo and parse the results (have written a small C# application for that). Works for 6 cams in my garden. The delay is about 30 sec. after upload. Only Telegram blocks sometimes because of to many uploads.... I ve not a good solution for it yet.

I also upload only single images on movement recognition. All my cameras are connected to Motion software so movement is detected by Motion and then the image(s) are sent to My Jetson Nano via MQTT for analys by YOLO v4. The YOLO analyze is runned in a python script and takes typically 0.3 second with my images 640x480. I can send many images to the Jetson Nano, they are put into a large queu if needed and then handled one-by-one. If a desired object (person n my case) is found, I get a thumbnail with the detected person in a notification via Telegram, the first is more or less instant if a person is detected. This has now worked flawless since a couple of years, migrated from YOLO v3, and is really good in detecting "unwanted visitis". I have added some additionla checks during the years to minimize false alarms like accepted ratios (width/height) that is expected from a human body etc etc. So far, good enough

2 Likes

PS If you are looking for real time streaming and analyze of video, the Jetson Nano is not enough, you need a much more powerful graphic card fro NVIDIA or similar, then the power consumption will also be something totally different and much higher

1 Like