search

cvat-ai / cvat

Annotate better with CVAT, the industry-leading data engine for machine learning. Used and trusted by teams at any scale, for data of any scale.
https://cvat.ai
MIT License
12.63k stars 3.01k forks source link

Segmentation using watershed #606

Closed HagegeR closed 9 months ago

HagegeR commented 5 years ago

First of all, very impressive work,

I wanted to know if you could implement a watershed segmentation option, a little bit like that:

https://github.com/abreheret/PixelAnnotationTool

the main backend function is simply watershed from opencv

it would be needed to add a "brushing" interaction with the canvas and a button to run the watershedding

nmanovic commented 5 years ago

Probably it can be a nice feature. Could you please estimate how much your annotation speed can be improved if the feature will be implemented?

HagegeR commented 5 years ago

well it will depend on the kind of annotation that will be done and on how accurate we want it to be, but I think it will be anywhere between 4 and 10 times faster on segmentation work.

hannes09 commented 3 years ago

Can you give an update about the status of this feature? Is there any chance to see a watershed tool in the near future?

DPeled commented 3 years ago

Hey, I have this feature implemented with python on my forked version of an older version of this repo. I can use it to write this feature using OpenCV.js. I'll take this on myself.

car0lin3d commented 3 years ago

Hey, I have this feature implemented with python on my forked version of an older version of this repo. I can use it to write this feature using OpenCV.js. I'll take this on myself.

Hey, I'm also interested in this feature. Did you start with the implementation already? Do you need support? I would also be interested in the python implementation. Do you have it online somewhere or would you share it?

DPeled commented 3 years ago

Hey, I have this feature implemented with python on my forked version of an older version of this repo. I can use it to write this feature using OpenCV.js. I'll take this on myself.

Hey, I'm also interested in this feature. Did you start with the implementation already? Do you need support? I would also be interested in the python implementation. Do you have it online somewhere or would you share it?

Hey @car0lin3d, yes I've started, but in slow progress. This is the first time I do thing in typescript so I need to gain some confidence.

I have the python implementation on github on this link (This is only the implementation of the watershed, but the transferring of the markers from the UI to the backend is located over the repo)

bsekachev commented 9 months ago

We have SAM now. It is much better than classic CV algorithms and we will integrate new. So, I do not consider the issue relevant anymore.