Open samirbr opened 4 years ago
Hey @samirbr,
Thanks for bringing that up! I think it can also be relevant for other labeling types, and the question we need to answer is how to put in a group of labels (possibly connected somehow) into the object (image).
<KyPointLabels name="kp">
<Label value="Nose" />
<Label value="Ear" />
<Label value="Lip" />
<LabelGroup value="Nose and Ear" connected="True" values="Nose,Ear" states="[[1,1], [2,2]" />
<LabelGroup value="Nose and Ear and Lip" connected="True" values="Nose,Ear,Lip" states="[[1,1], [2,2], [3,3]]" connections="[[0,1],[0,2]]" />
<KeyPointLabels>
That would provide labels that can create 2 & 3 keypoints with the initial state being the location of the keypoints on the image. And for three keypoint label nodes are going to be connected using the line. I think it's a general approach that would enable you to create multiple elements at once, and also extendable further if we introduce an ability to ask for the initial state from the user/model.
Overall it's in the pipeline, but need to think more about implementation, ideas are more than welcome.
Hey,
I need to annotate video using keypoints. I am also looking for solution like this. Can somebody please explain to me about this. I want to make a skeleton structure using 100 keypoints.
Hi, LS doesn't support skeleton structure with keypoints yet, but you can try keypoints grouping using the region panel. https://www.youtube.com/watch?v=V3ieJGky3yc
Hey,
Does label-studio support "keypoints grouping using the region panel" using Videos??? I have lots of videos and I have to start with 20fps.
Thanks and Regards, Pooja Kumari
On Mon, May 9, 2022 at 11:50 PM Max Tkachenko @.***> wrote:
Hi, LS doesn't support skeleton structure with keypoints yet, but you can try keypoints grouping using the region panel. https://www.youtube.com/watch?v=V3ieJGky3yc
— Reply to this email directly, view it on GitHub https://github.com/heartexlabs/label-studio/issues/141#issuecomment-1121617702, or unsubscribe https://github.com/notifications/unsubscribe-auth/AQB6DY52Z4KSLC4EIMGYQLLVJGCA3ANCNFSM4JVNTDMQ . You are receiving this because you commented.Message ID: @.***>
In video labeling you can use only bounding boxes for now. Yes, you will be able group bboxes using the region panel.
Is your feature request related to a problem? Please describe. There is a case where the Key Points might be related (i.e. same person
Nose
,Hand
). It would be easier to relate different Key Points to a same "entity" (i.e. same person) if they are connected.Describe the solution you'd like Add a tag to enable Key Point connection (i.e.KP Node to KP Ear) and Key Point connection schema.
Possible implementation?