Multi-input models support? (Zeroshot Detection)

I’m trying to get Zero shot detection to work with Hailo8, specifically YOLO-World (AILab-CVC/YOLO-World) model.

I’ve been following the Dataflow compiler tutorials with jupyter notebook, but it’s hard to get a grasp on how to optimize, especially making nms config files an alls configuration files.

  1. Is there a similar model structure or a good place for me to start?
  2. After compilation, how do I give text inputs for inference, considering it’s a YOLO detection model with text (open-vocabulary detection capabilities)?


Hey @developer,

Welcome to the Hailo Community!

You can find the full guidelines for the Dataflow Compiler (DFC) in the Hailo Developer Zone Documentation.
This documentation covers everything you need to know about using the DFC.

Regarding optimizing the YOLO-World model, we currently do not have that specific model available in our Model Zoo. However, we do have a zero-shot segmentation model in the Model Zoo that you can use for a CLIP application.

Please let me know if you have any other questions! I’m happy to help.

Best regards,