Export Annotated Images with Overlaid Annotations

[lekhamadhavan]

Description

It would be very helpful to have an option in CVAT to export annotated images with annotations visually overlaid on the original/raw images.

Currently, exported datasets contain annotation files separately (COCO, YOLO etc.), but there is no straightforward option to generate preview images showing:

Bounding boxes
Polygons
Labels
Segmentation masks

directly rendered on top of the original images.

Use Case

This feature would help with:

Quick visual verification of annotations
QA/review workflows
Sharing annotated previews with clients or non-technical stakeholders
Debugging annotation issues before training

Suggested Feature

Add an export option such as:

"Export annotated preview images"
"Render annotations on images"

Possible output formats:

ZIP containing rendered images
PNG/JPG preview images
Optional configurable colors/labels visibility

Additional Notes

This would significantly improve annotation validation workflows and reduce the need for external visualization scripts/tools.

[bsekachev]

Did you try format "Segmentation masks 1.1". It does exactly what you describe.

[lekhamadhavan]

Thanks for the suggestion. I tried the "Segmentation masks 1.1" export format, but it generates only the mask image where the background and masks are rendered as solid colors.

What I am looking for is slightly different:

• Export the original/raw image with annotations visually overlaid on top
• Similar to how annotations appear inside the CVAT UI
• Including polygons/bounding boxes/labels/masks rendered on the actual image

The main purpose is visual verification and QA while still keeping the actual dataset export separately in formats like Datumaro or YOLO Segmentation for downstream ML tasks.

So ideally, the workflow would be:

1. Export dataset in Datumaro/YOLO Segmentation format for training
2. Additionally export preview images with rendered annotations for manual validation/review

Currently, this requires external scripts/tools for visualization, so having native support in CVAT would be very useful.

[Stive1999]

hello everyone I am Stive and l found there is no project going on at the moment ,, can someone help me to find any ongoing project to provide me with the link please