Skip to content

Florence2 produces empty grounding results #26

@Impulse-MkV

Description

@Impulse-MkV

Hi, thanks for your great work.

I encountered some problems when I am trying to reproduce the XVerseBench.

When I run the assets/segmentation.py, I found the detector.get_multiple_instances (line 34) produces empty results on all images in the DreamBench+ dataset.

More extactly, I found this line in grounded_sam2_florence2_autolabel_pipeline.py (line 255) produces empty results.

grounding_results = self.run_florence2('<OPEN_VOCABULARY_DETECTION>', text_input, image)['<OPEN_VOCABULARY_DETECTION>']
Then the program always goes to the if branch in line 257-258.

I download the checkpoints of Florence2 using the provided download_ckpts script. However, when I load in this weight, the code returns such a warning:

Some weights of Florence2ForConditionalGeneration were not initialized from the model checkpoint at ./checkpoints/Florence-2-large and are newly initialized: ['language_model.lm_head.weight'] You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.

I'm not sure whether it is correct. I would be very grateful if you can help me solve this problem. Thanks for your help in advance.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions