Just a quick question:
I have a Blink outdoor camera with a security light.
Ideally, I need to program it so as not to keep going off and using the battery.
We live remotely so it’s not going to disturb the neighbors.
A library item in TensorFlow with a similar action (recognising people and animals, sheep, rabbits, the postie, the occasional farm hands, our cat, etc)
Additionally, recognises the wind which can be pretty powerful and it doesn’t keep pinging me or the light going off.
A good addition may also be a recorded voice to warn people are trespassing. That would be brilliant.
I may need a step-by-step approach as although I can edit TensorFlow js, I am not confident in producing this all from scratch.
Any links would be a great help.
So I do not know anything about Blink but assuming you can stream that to some device you own you could follow my tutorial here to make your own smart cam:
Thank you so much! I will take a look very soon as in the midst of moving. Blink is just a centralised app where I can control all the blink cameras (a trading name). https://www.practicalsmarthome.com/a-guide-to-blink-smart-cameras-and-systems/ I guess this will explain.
Just one more thing, I took your class online (the google course)
it was very very good!!! thank you. I just wanted to go one step further and include the warning and the recognising of animals and certain faces to not enable a warning.
So the COCO-SSD model I use should recognize some animals like cats and dogs too. However for better reocgnition you can train your own object detection model based on YoLOv7 or similar as explained in this tutorial
or his more advanced version and recent one here:
With all of the above you could then make a TFJS Model that runs in browser to recognize anything you want.