batch more than 1

General Discussion
21

Sure, this is the json file

{
    "Checksum": "3314f4fcd9099b9085ee79d010affe0c3975ab2be811d26fca7cf252c92e6239",
    "ConfigVersion": 1,
    "DEVICE": [
        {
            "DeviceType": "HAILO8",
            "RuntimeAgent": "HAILORT",
            "SupportedDeviceTypes": "HAILORT/HAILO8",
            "ThreadPackSize": 6
        }
    ],
    "PRE_PROCESS": [
        {
            "InputType": "Tensor",
            "InputN": 1,
            "InputH": 928,
            "InputW": 928,
            "InputC": 3,                
            "InputRawDataType": "DG_UINT8"
        }
    ],
    "MODEL_PARAMETERS": [
        {
            "ModelPath": "yolov8m.hef"
        }
    ],
    "POST_PROCESS": [
        {
            "OutputPostprocessType": "None"
        }
    ]
}
22

If you would change "InputType": "Tensor" to "InputType": "Image" then you may benefit from PySDK pre-processing capabilities like resizing with letterboxing etc. Yuo may feed your camera output directly to the model: all the resizing and BGR->RGB conversion will be taken care by PySDK preprocessor. And model_time_profile will work as well :slight_smile:

23

Thanks a lot! Now it works well!

1 Like