Detic 2D object detector

[Dcasadoherraez]

Hello!
Thanks so much for your contribution with RegionPLC.
I am trying to run it on a custom dataset so I need to generate the Detic 2D detections myself.
Is there any specific procedure I should follow? I couldn't find any instructions in the RegionPLC repo about this.
I noticed I have to provide the class names to the detector according to their repo

Features
Detects any class given class names (using [CLIP](https://github.com/openai/CLIP)).

Is this correct?

I have seen that caption_kosmos_and_detic_iou0.2-0.0.json also looks like:

    "samples_LIDAR_TOP_n015-2018-07-18-11-07-57+0800__LIDAR_TOP__1531883534449471": {
        "cam_front_0": "A truck",
        "cam_front_1": "several other trucks",
        "cam_front_2": "a few people",
        "cam_front_left_0": "several trees",
        "cam_front_left_1": "several trees",
        "cam_front_right_0": "A man",
        "cam_front_right_1": "a large white truck",
        "cam_front_right_2": "two cars",
        "cam_back_0": "a car",
        "cam_back_1": "a road",
        "cam_back_2": "a building",
        "cam_back_3": "a traffic light",
        "cam_back_4": "trees",
        "cam_back_5": "the sky",
        "cam_back_6": "two cars",
        "cam_back_left_0": "two people",
        "cam_back_left_1": "two people",
        "cam_back_left_2": "a building",
        "cam_back_left_3": "trees",
        "cam_back_right_0": "The truck",
        "app_0_0": "fireplug",
        "app_0_1": "person",
        "app_0_2": "streetlight",
        "app_0_3": "signboard",
        "app_0_4": "headlight",
        "app_0_5": "poster",
        "app_0_6": "signboard",
........

Can you please share the script to combine the script from Detic to output the required format?

Thanks in advance :)