DLTA-AI : Data Labeling, Tracking and Annotation with AI

[0ssamaak0]

Based on the amazing library of mmdetection models, DLTA-AI provides a fully automated workflow for image and video annotation, with support for both segmentation and tracking, and zero shot segmetnation and tracking as well (with SAM)

I hope to hear your feedback and criticism

Repo Link
Video

[hhaAndroid]

@0ssamaak0 Thank you very much. Can you tell me what is the difference between DLTA-AI and labelstudio? We have made a simple demo, what should we do if we want to connect to DLTA-AI? Thank you!

[0ssamaak0]

Regarding the differences:

I think the main difference is the customizability, user can choose from (+50) models for segmentation tasks, export formats can be user-defined too
Beside this:

• the video mode is very user friendly and convenient
• the integration of SAM to DLTA-AI provides many options, like: tracking with SAM, and enhancing the segmentations of other models with SAM

I really invite you to check the README for more details, and try it yourself

Regarding the second question: I didn't get it, what do you mean by: connecting to DLTA-AI?

Thank you 😃

[0ssamaak0]

If you mean by connecting: using mmdetection API, it's already implemented, the Model Explorer provides almost all mmdetection v2 models to select from

[hhaAndroid]

If you mean by connecting: using mmdetection API, it's already implemented, the Model Explorer provides almost all mmdetection v2 models to select from

OK.