Custom HEF on Raspberry Pi 5: “No tensor with name resnet_v1_50/softmax1” error in rpicam-hello

General Discussion
, , , ,
1

Hello Hailo Team,

I am trying to perform real-time inference using a Raspberry Pi 5, Hailo-8, and an IMX708 camera.

I have already fine-tuned a MobileNet-based model and successfully converted it into a .hef file.

When I run the example pipelines provided in the Hailo Apps infrastructure, everything works correctly.

For example, the following command runs without any issues:
rpicam-hello -t 0 --post-process-file /usr/share/rpi-camera-assets/hailo_classifier.json

However, when I try to run my custom model, I encounter the following error:
what(): No tensor with name resnet_v1_50/softmax1
Aborted

The command I used is shown below:
rpicam-hello -t 0 --post-process-file /usr/share/rpi-camera-assets/my_inference.json

The contents of the my_inference.json file are as follows:

{
    "rpicam-apps":
    {
        "lores":
        {
            "width": 224,
            "height": 224,
            "format": "rgb"
        }
    },

    "hailo_classifier":
    {
        "hef_file": "/home/mataflow/hailo_models/best_for_rpicam.hef"
    },

    "annotate_cv" :
    {
        "text": "",
        "fg" : 255,
        "bg" : 0,
        "scale" : 1.0,
        "thickness" : 2,
        "alpha" : 0.3
    }
}

Could you please advise how to resolve this issue?

Thank you in advance.