Number of annotations created by Label Assist is limited to 300?

Hi,

I’m experiencing issues while using Label Assist. I have images where I need to segment object instances, with the number of objects typically ranging from 282 to 956. Labeling these images manually is very time-consuming. To address this, I trained a model on a smaller subset of the data—just a few images—and then uploaded this model to use with Label Assist to save time.

When I use the trained model offline for predictions, I can set the maximum number of detections. However, I haven’t found a way to do this within Label Assist. For some reason, it always labels exactly 300 objects, even when there are around 800 objects in the images. When I run predictions with the same model offline, it correctly detects more than 300 objects. This suggests there might be a maximum detection limit when using Label Assist.

Is there any way to adjust this limit or work around it?

Project type: Instance segmentation
Browser: Google Chrome
Model: YOLOv11

Thank you for your time and assistance.

Hi there,

There isn’t currently a way to set a higher MAX_OBJECTS param in Label Assist, but I’ve surfaced this with the team to add to our roadmap!

The best workaround I can think of for now, is using our API to get annotation batches, pull images, run your model locally, and then use the API to upload annotations back to Roboflow: Batches | Roboflow Docs. I hope this helps!

1 Like