Advice on creating an own dataset for segmentation

Deep Learning
1

Hello,

I would like to build a dataset which I would like to annotate in detail.
I am wondering whether some people here have experiences with good software tools to do so.
Preferably software tools who are a bit user friendly and allow to annotate images to use later for semantic segmentation.

I am not sure if it is important but the images are medical, tif and grayscale.

I have fiddled a bit with QuPath, but to me this was far from the easiest software to use and I it was not clear to me how I could use it well to get good annotations and complete (annotating in detail was not really a problem).

Any advice is appreciated.
Louis

2

Ha yes I was going to recommend QuPath… it seems to work reasonably well. ImageJ comes to mind. On the paid side of things I have heard of inForm (Akoya)… haven’t used that one myself.

3

Have you tried the VIA tool?
Google VIA annotation tool. That’s what we use in my company. Works for us for now

4

Oh its free and open source too

5

Ah cool. I’ll check it. I have been playing around with APEER which has a free version and it seems to work pretty well as well.

VIA seems pretty light weight which is cool.

1 Like
6

Nice I’ll check it out.
We use VIA to annotate then we have a bunch of python scripts that generate the segmentation. If you need help i can give you a few examples. DM