Open sdranju opened 3 months ago
You can determine the size of the image during inference by modifying the parameter target_size, and the image will be resized to the size you set.
Thank you for your response. I'm asking because I'll feed the frame directly from the camera using Camera2 or CameraX api. I need real-time detection, resizing may create extra overhead to the system and lag.
Is the model accept dynamic input shape or do I need to resize the image to a specific size required by the model?