search

dbolya / yolact

A simple, fully convolutional model for real-time instance segmentation.
MIT License
5k stars 1.33k forks source link

Result data #502

Open hyunaseo opened 4 years ago

hyunaseo commented 4 years ago

Thanks for your work so much! I have two questions!

  1. I want to work with the predicted result from video, like bbox, class name, segmentation area, etc. And how can I get these results like json file or any other types?

  2. If there are two objects in the same category and if they are little bit overlapped, the yolact result can record it with recognizing those two objects seperately?

I'll be glad if you give me some answer! Thanks a lot🤩

MistaFrowns commented 4 years ago

I'm working a similar issue; is there a way to get a shapefile of the objects?

In going through the code for 'eval.py'...

'masks' = a mask for each detected object 'boxes' = coordinates for each bounding box

Is there a set of polygons somewhere in the code, or do I need to create that using an external algorithm on the mask?

hyunaseo commented 4 years ago

@MistaFrowns Hello, Well I'll try to open the codes more closely today, so if there is anything I find out, then I will write a comment! Let's try :-)

areebsyed commented 3 years ago

@hyunas1996 @MistaFrowns did you find out anything about the polygon info. How to extract the polygon points from the code?