Confusing Bounding Box detectnet_v2

I have succeeded create model from custom training dataset to detect person from different camera angle. But,when I try to see its inference its resulted in too many bounding box. Is it normal?

https://drive.google.com/file/d/1eePtoHimaLi2NtzO2tgcEgdv1KFUCN1O/view?usp=sharing

How about the tlt-evaluate result or the validation AP(mAP) during your training?

This is model validation result during training:

Validation cost: 0.000388
Mean average_precision (in %): 66.6667

class name      average precision (in %)
------------  --------------------------
person                           66.6667

Median Inference Time: 0.043674

And this is model evaluation result:

Validation cost: 0.003988
Mean average_precision (in %): 75.0000

class name      average precision (in %)
------------  --------------------------
person                                75

Median Inference Time: 1.582642

So, which tool did you use to do inference? Is it tlt-infer?

I used tlt-infer, using this command:

# Running inference for detection on n images
!tlt-infer detectnet_v2 -i $DATA_DIR/testing_images/ \
                        -o $USER_EXPERIMENT_DIR/tlt_infer_testing \
                        -m $USER_EXPERIMENT_DIR/experiment_dir_retrain/weights/resnet18_detector_pruned.tlt \
                        -cp $SPECS_DIR/detectnet_v2_clusterfile_kitti_pc.json \
                        -k $KEY \
                        --kitti_dump \
                        -lw 3 \
                        -g 0 \
                        -bs 64

Could you please paste the detectnet_v2_clusterfile_kitti_pc.json?

{
    "dbscan_criterion": "IOU",
    "dbscan_eps": {
        "person": 0.35,
        "default": 0.15
    },
    "dbscan_min_samples": {
        "person": 0.05,
        "default": 0.0
    },
    "min_cov_to_cluster": {
        "pedestrian": 0.005,
        "default": 0.005
    },
    "min_obj_height": {
        "pedestrian": 4,
        "default": 2
    },
    "target_classes": ["person"],
    "confidence_th": {
        "person": 0.6
    },
    "confidence_model": {
        "person": { "kind": "aggregate_cov"},
        "default": { "kind": "aggregate_cov"}
    },
    "output_map": {
        "person" : "person"
    },
    "color": {
        "person": "cyan",
        "default": "blue"
    },
    "postproc_classes": ["person"],
    "image_height": 384,
    "image_width": 1248,
    "stride": 16
}

Please modify the “pedestrian” to “person”.

Thank you for your fast response