gui for dataset creation and annotation export

[GallonDeng]

for the Collaborative Mode of annotation, is there a simple way to create a unlabeled dataset from a local directory and then export the annotations, not use the python code in the terminal ?

[imhuwq]

Hi @AllenDun, setting up a local directory as a database is straightforward. You can find step-by-step instructions in our documentation.
And there is a GUI way to export annotations already.
If you have any questions or need further help, don't hesitate to reach out!

[GallonDeng]

Hi, @imhuwq thanks for your reply. Yes, it can be found in the documentation for how to set up a local directory as a database, to use python code, and I tried. There is also a GUI way to export annotations, but it is for quick mode not for Collaborative Mode of annotation. It was not so straightforward for collaborative mode to create a unlabeled dataset and export the annotation result since I tried. Anything I missed? I installed the dds with pypi.

[imhuwq]

Hi @AllenDun, I apologize for the confusion earlier regarding exporting datasets and annotations; currently, we only support exporting annotations to datasets.
We don't have plans to add dataset exporting to files immediately, but we value community feedback and may reconsider based on user needs.
Could you briefly describe your use of DDS, especially how you handle labeling and exporting?

[GallonDeng]

DDS is a tool for Visualization, Annotation, and Model Analysis, which is good. But for a complete loop, we usually upload a image directory, do labeling, do visualization or analysis, and export the final correct annotations for model training. This process sometimes is done by people without code experiences.

[imhuwq]

I'll discuss this feature with my team and will inform you as soon as we reach a decision.

[imhuwq]

Hi @AllenDun, great news!

We've decided to go ahead with the feature of exporting datasets to local files.
Right now, we're in the thick of designing the GUI and figuring out the best data format to use.

We'd love to get your input on what you're expecting and any specific requirements you have in mind.
Looking forward to your thoughts!

[GallonDeng]

glad to hear that. for the data format, at least support COCO format. These days, I continued to try DDS, it seems that intelligent interactive annotation only work for rectangle label with iVP for the local version now. In the doc, it says the feature is not available in the local version. Intelligent annotation is one of the most import feature for many new annotation tools, e.g., X-AnyLabeling, anylabeling, even cvat. It would be great if DDS could release the feature for local version.

[imhuwq]

Hi @AllenDun, thank you for your invaluable feedback.

1. COCO Format
   Absolutely, considering its popularity, COCO format is a top choice for us. Your endorsement further boosts our confidence in this decision.

2. Intelligent Annotation
   We've just released support for annotating and saving masks, polygons, and keypoints in V0.10.0.
   But the intelligent annotation in the opensource DDS tool has it's limitation. The DDS tool doesn't install models and algorithms on the user's computer. Instead, it calls our API platform with a public token behind, subject to strict rate limits.
   We're planning to allow users to use their own API tokens in the DDS tool, providing individual quotas and rate limits, and making intelligent annotation ready for production use.
   If you're interested in scaling up intelligent annotation with the DDS tool, please don't hesitate to email us at weiliu@idea.edu.cn to apply for an individual token for free.