Feature request: get objects after they have been created

[frndmg]

My actions before raising this issue

• Read/searched the docs
• Searched past issues

I am trying to create a set of points and group them using the reducers createAnnotationsAsync and groupAnnotationsAsync in the following way:

onCreateAnnotationsAndGrouping(sessionInstance: any, frame: number, states: any[]): Promise<void> {
    return dispatch(createAnnotationsAsync(sessionInstance, frame, states))
       .then(() => dispatch(groupAnnotationsAsync(sessionInstance, frame, states)))
}

When executing groupAnnotationsAsync the following error is displayed:

Error: The object has not been saved yet. Call annotations.put([state]) before

I have tried setting a timeout before grouping takes place but the same error appears.

The grouping tool works well when I do it by hand.

Expected Behaviour

It should group the points and do not throw any errors.

Current Behaviour

An error is thrown and no grouping occurs.

Possible Solution

I think it should be a bug on the cavt-core API that is causing this. But I can't be sure.

Steps to Reproduce (for bugs)

The code is already shown.

Context

I am trying to develop a way to annotate human pose using keypoints. Right now the approach can lead to pitfalls because you are not forced or encouraged to use correct labels in your annotation process. For now, I just want to programmatically add this points and group them in order to have a template for the human figure.

Eventually, I would like to add a template designer and create a model on the backend side.

It will be good to create a new shape that assembles these graph structures as well.

Your Environment

• Git hash commit (git log -1): 81a13cb
• Docker version docker version (e.g. Docker 17.0.05): Not using docker for the development environment but, 18.09.5
• Are you using Docker Swarm or Kubernetes? No
• Operating System and version (e.g. Linux, Windows,): uname -srvo: Linux 4.15.0-74-generic #84-Ubuntu SMP Thu Dec 19 08:06:28 UTC 2019 GNU/Linux

Next steps

I already asked this question on Gitter

[frndmg]

You can try this https://github.com/signatrix/cvat/tree/0c85aa1fc2ddfafafc1ff82b2e2753f44c726da6

[bsekachev]

@frndmg

In your example you pass states to groupAnnotationsAsync that weren't saved in collection (message is about it). It's wrong using of cvat-core API.

Calling of job.annotations.put(states) (inside createAnnotationsAsync) returns states saved in collection. You should group them instead of.

[frndmg]

@bsekachev

Thanks a lot!

I created a new reducer on annotation-actions.ts. I was not sure how to get the new states out of the createAnnotationsAsync.

[bsekachev]

Theoretically you can get the latest data from the redux storage after createAnnotationsAsync was executed.
But creating a new action and a reducer (as you suggested) is better way I think.

Update:
Also I realized I was mistaked.
sessionInstance.annotations.put doesn't return anything. Just need to get data from the core using sessionInstance.annotations.get after you have done put.

[frndmg]

How could I retrieve the points I just added? @bsekachev

[bsekachev]

@frndmg

Hmm..
I've just checked how it is possible and I haven't found nice way to do it.
The best option would be to return some information about added objects from annotations.put I suppose. For that we need to update API of cvat-core.

[bsekachev]

@frndmg

With the PR above you can do something like:

const ids = await session.annotations.put(objectsToSave);
const allObjects = await session.annotations.get(frame);
const objectsJustAdded = allStates.filter((state) => ids.includes(state.clientID));

[frndmg]

I see. This requires fewer changes than I was doing.

I assume it should be allObjects on the last line.

I will add that piece of code until the PR is merged, thanks!

[bsekachev]

I assume it should be allObjects on the last line.

Sure.
Not at all