Multiple versions of .hef converted from my custom .onnx model but none compatible with Rpi5 +AI HAT+(hailo8l)

One more thing, I was checking the coco90 label and noticed they have got index 0 as “unlabeled”. Do I have to consider this in my labels.json? such as in adding “unlabeled” before “echinus”?

* Copyright (c) 2021-2022 Hailo Technologies Ltd. All rights reserved.
* Distributed under the LGPL license (https://www.gnu.org/licenses/old-licenses/lgpl-2.1.txt)
**/
#pragma once
#include <map>
namespace common
{
    static std::map<uint8_t, std::string> coco_ninety_classes = {
        {0, "unlabeled"},
        {1, "person"},
        {10, "traffic light"},
        {11, "fire hydrant"},
        {13, "stop sign"},
        {14, "parking meter"},

Hi @Liping_Jin,

Yes - add a placeholder at index 0.** The decoder maps model class N → label index N+1 (that’s why your echinus/class-0 showed as “person”/COCO index 1). So:

{
    "labels": ["unlabeled", "echinus", "holothurian", "scallop", "starfish"]
}

Without it, all your labels shift by one. The index-0 text can be anything (“unlabeled”, “background”) - it’s never displayed.

On syntax docs: not in hailo --help. The valid JSON keys come from the rpicam-apps stage source - check the Read() method in post_processing_stages/hailo/hailo_yolo_inference.cpp, plus the Raspberry Pi camera docs: https://www.raspberrypi.com/documentation/computers/camera_software.html

Thanks,

1 Like

Thank you so much. :heart_suit: