wkentaro / labelme

Image Polygonal Annotation with Python (polygon, rectangle, circle, line, point and image-level flag annotation).
https://labelme.io
Other
13.56k stars 3.41k forks source link

Use labelme with iris dataset and yolo darknet #367

Closed abdou31 closed 5 years ago

abdou31 commented 5 years ago

Hello Sir, I want to know how can i use this tool with iris dataset (custom dataset ) and also with yolo darknet? Thanks

wkentaro commented 5 years ago

I don’t know much about iris dataset (is it a flower dataset?). But what you need to do is annotate image with polygon or bounding box.

To use annotated data, you need to create a script equivalent to labelme2voc.py and labelme2coco.py (e.g., labelme2darknet.py).

abdou31 commented 5 years ago

Iris data means the dataset of iris region in the eye , i worked on this to loacte the iris region and i use darknet yolo forked by AlexayAB , the annotation of dataset e.g CASIA with four values giving box , I don't need drawing box on iris , i need to draw the form of the region of iris ( not always circle ) , so for that reason , i choose to use the segmentation method to do that

wkentaro commented 5 years ago

Ok then, what is you're question?

abdou31 commented 5 years ago

My question is : Is it possible to labelling iris data using Labelme tool ? Do you know how to do that? ( just give an exemple )

wkentaro commented 5 years ago

Is it possible to labelling iris data using Labelme tool ?

You can just annotate an image with polygons (or bounding boxes), right? As demonstrated in README: https://github.com/wkentaro/labelme#description.

Why do you think it may be not possible?

abdou31 commented 5 years ago

For example this image: I want to annotate the iris region in the eye using segmentation , did you understand me now? https://i.imgur.com/YBxcg0s.jpg

wkentaro commented 5 years ago

Yeah. You can annotate iris with polygon annotations.

Actually, I'm still not sure what you're asking..

abdou31 commented 5 years ago

My question is clear , that is how to do segmentation with object or with iris region ?

wkentaro commented 5 years ago

Are you asking about how to use machine learning method for object segmentation?

If so, you should read paper or ask people on forum. Here is not the place it supposed to be. This repo is about human manual annotation, not how to use learning method for segmentation.

abdou31 commented 5 years ago

Sorry , I mean annotate segmentation with iris regions ?

wkentaro commented 5 years ago

You want manually annotate the region of an iris in an image?

If so, you can use the polygon annotation of labelme as I said before.

abdou31 commented 5 years ago

yes using segmentation ( coloring the region of iris ) with your tool

wkentaro commented 5 years ago

Ok. Please use polygonal annotation of this tool.

abdou31 commented 5 years ago

Ok thanks , can you give me instruction beacause i dont find segmentation option?

wkentaro commented 5 years ago

Have you read the tutorial? https://github.com/wkentaro/labelme/blob/master/examples/tutorial/README.md

abdou31 commented 5 years ago

I read just the main page of the github tool's

wkentaro commented 5 years ago

I’m just not sure why you need instruction.

wkentaro commented 5 years ago

I don’t have anything to explain more aside from the readme.

abdou31 commented 5 years ago

I need instruction because i don't understand how to use segmentation to labelling the region of iris (images)

wkentaro commented 5 years ago

What have you tried and what you don’t understand about this tool?

abdou31 commented 5 years ago

I have tried using keypoints to labelling image and i dont understand how to use this segmentation after that , Note: I will read the instructions Thanks ,

wkentaro commented 5 years ago

It's out of the scope of this repo, right? This repo is for annotations not how-to-use-annotations.

Maybe you should check state-of-the-art segmentation or detection system (e.g., YOLO, PSPNet, Mask R-CNN etc.), and check how to use polygonal annotation on that system. There are some common formats of export (e.g., COCO, VOC) that are already included in this repo. So if you can use it, use it. If not, maybe this repo is not helpful for you.