Feature Requests

My dataset images are stored on S3. I'd like to import them to LabelFlow by linking my S3 bucket directly.

Since different teams in Operations have different roles. The access of labelflow to each user should be different. For example : an outsourcing team to labelise the images should not be allowed to import or export image files or have access to any other image file

The best visualization tool. Please add some feature the same feature in https://www.robots.ox.ac.uk/~vgg/software/via/via.html . Detail: With each object add attribute option: Textbox, Dropdown, Checkbox, Boolean, Description for each class, Relation between components. I expect this feature from yr team soon. Thank you!

As you have likely observed, creating a new workspace can be relatively slow (~10 sec). This is due to the tutorial dataset which is created together with every new workspaces. We have several ideas to make this process much faster in a near future.

To label the Key-value problem. Please add feature: object link to object another corresponds to object key link to object value. As in the picture. I look forward to hearing from yr team

When creating a dataset, it's important to have a place to share labeling instructions, write down informations about the data (origin, quality, etc.), leave notes to colleagues about the annotation work, save bookmarks etc. We will be adding a tab (Overview) inside every datasets, you'll be finding: A space where you can write some notes (markdown) A space where you'll find general statistics A space with class balance informations

On the labeling view, add a button to allow users to easily print an image with its labels.

Workflows would allow to split the work between different people or different teams: annotators, Quality checkers, dataset owners etc.

May you update with DICOM annotator tools. Or just an upload DICOM with automatic convert to jpg/png image.

Thanks you Nicolas for sharing the link for comments regarding the Labelflow. We have tested labelflow and below is the reply from my team We have gone through the annotations, it seems that they have used bounding boxes for annotation. However Object Identification wont work and we would need NER annotations which are done after extracting text. Please note NER annotations can be in JSON or IOB format (Rational Index Format for NER). Please let me know if we can do NER Annotation in Labelflow. What we required; We grab the text from images and save into a TXT file. While we upload TXT to labelflow , we want to do NER annotation with specified classes. Please advise.