-1

Is it possible to detect exact coordinates (pixels) of an object in Python?

Currently i'm working with ImageAI and YOLOv3 models, but it provides only a box with the object. I haven't found any references in ImageAI docs how to detect exact coords.

My project should detect bottles on pictures with different backgrounds and fetch those bottles. Due to unstable backgrounds, it's really difficult to remove such backgrounds, so it could be great to fetch exact bottle coordinates and "copy" them from the original picture.

I've read something about tensorflow, but it's quite difficult to get into to, so if anyone can give some information/directions about such libraries and models (especially for objects like bottle), i would be really happy.

Nokt
  • 29
  • 5
  • I think this question is too general and unfocussed for Stackoverflow. Also, you don't define what you mean by "exact coordinates". What are the "exact" co-ordinates of a bottle in an image? Can you give an example image and the desired result? It's not clear to me. – Bill Aug 26 '23 at 17:21
  • Consider deleting this question and re-posting it here: [Artificial Intelligence Stack Exchange - questions tagged image-recognition](https://ai.stackexchange.com/questions/tagged/image-recognition) – Bill Aug 26 '23 at 17:26
  • Take a look at image segmentation and u-nets. They can perform pixel-based detection instead of bounding boxes. – mhenning Aug 26 '23 at 21:00

0 Answers0