search

gerrito333 / mousehunter-edge

Cat with prey detection on Raspberry Pi. Lock cat pet flap if prey is detected. Object detection implemented in TFLite with ImageNet v1 SSD. Inference on EdgeTPU (Google Coral USB). Stores images on AWS S3 and sends notifications to iOS device.
Other
72 stars 18 forks source link

How did you labeled the images? Whole image or "just" the relevant part? #4

Open jojowe opened 10 months ago

jojowe commented 10 months ago

Hi @gerrito333,

I forked your mousehunter-edge and I'm customizing the source code to my needs (e.g. using telegram for notification).

I have a question which aims to the labeling of the images... Had you labeled the whole image (prey or no prey) or "just" the part (cats snout with and without prey) to prepare the data for the model training?

For my understanding, the labeling of the whole image is detailed enough, because it is "only" a classification and not a object detection problem.

Thanks in afford for your reply.

Best regards