Custom resnet50 classification model secondary-gie0 outputs only first entry in label.txt

[17396156679]

Please provide complete information as applicable to your setup.

**• Hardware Platform (Jetson / GPU) Tesla T4
**• DeepStream Version deepstream5.0.1
**• TensorRT Version 7.0
**• NVIDIA GPU Driver Version (valid for GPU only) 440.118.02(10.2)

I print some output inside the parsing function
parse function:

print output:

I found that the probability value output in outputCoverageBuffer is wrong, causing a wrong result .

PGIE config info:
[primary-gie]
enable=1
gpu-id=0
model-engine-file=…/engine/yolov5l_helmet.engine
labelfile-path=…/label/helmet_labels.txt
batch-size=5
#Required by the app for OSD, not a plugin property
bbox-border-color0=1;0;0;1
bbox-border-color1=0;1;1;1
bbox-border-color2=0;0;1;1
bbox-border-color3=0;1;0;1
interval=10
gie-unique-id=1
nvbuf-memory-type=0
config-file=config_infer_primary_yoloV5_Helmet.txt

[property]
gpu-id=0
net-scale-factor=0.0039215697906911373
#0=RGB, 1=BGR
model-color-format=0

custom-network-config=yolov3-tiny.cfg

model-file=yolov3-tiny.weights

model-engine-file=…/engine/yolov5l_helmet.engine
labelfile-path=…/label/helmet_labels.txt
process-mode=1

0=FP32, 1=INT8, 2=FP16 mode

network-mode=2
num-detected-classes=2
gie-unique-id=1
network-type=0
#is-classifier=0
output-blob-names=prob

0=Group Rectangles, 1=DBSCAN, 2=NMS, 3= DBSCAN+NMS Hybrid, 4 = None(No clustering)

output-tensor-meta=1
cluster-mode=2
maintain-aspect-ratio=1
parse-bbox-func-name=NvDsInferParseCustomYoloV5
custom-lib-path=…/nvdsinfer_custom_impl_Yolo/libnvdsinfer_custom_impl_Yolo_helmet.so
engine-create-func-name=NvDsInferYoloCudaEngineGet
#scaling-filter=0
#scaling-compute-hw=0

[class-attrs-all]
nms-iou-threshold=0.5
pre-cluster-threshold=0.4

PGIE labels.txt:
helmet
no_helmet

SGIE0 config info:
[secondary-gie0]
enable=1
gpu-id=0
model-engine-file=…/engine/resnet50_secondary0_helmet.engine
labelfile-path=…/label/helmet_secondary0_labels.txt
gie-unique-id=2
operate-on-gie-id=1
config-file=config_infer_secondary0_resnet50_Helmet.txt
operate-on-class-ids=0
batch-size=5
bbox-border-color0=0;0;1;1
bbox-border-color1=0;1;1;1
bbox-border-color2=0;0;1;1
bbox-border-color3=0;1;0;1

[property]
gpu-id=0
net-scale-factor=1
offsets=103.939;116.779;123.68
#offsets=124;117;104
#net-scale-factor=0.0039215697906911373
#0=RGB, 1=BGR
model-color-format=0
model-engine-file=…/engine/resnet50_secondary0_helmet.engine
labelfile-path=…/label/helmet_secondary0_labels.txt
#1:pgie 2:sgie
process-mode=2

0=FP32, 1=INT8, 2=FP16 mode

network-mode=2
num-detected-classes=3
gie-unique-id=2
operate-on-gie-id=1
operate-on-class-ids=0
batch-size=1
#0: Detector,1: Classifier,2: Segmentation
network-type=1
input-object-min-width=64
input-object-min-height=64
output-blob-names=output1;output2
classifier-async-mode=1
classifier-threshold=0.5
secondary-reinfer-interval=0

0=Group Rectangles, 1=DBSCAN, 2=NMS, 3= DBSCAN+NMS Hybrid, 4 = None(No clustering)

#output-tensor-meta=1
cluster-mode=2
maintain-aspect-ratio=1
parse-classifier-func-name=NvDsInferClassiferParseCustomResnet50_SGIE0
custom-lib-path=/opt/nvidia/deepstream/deepstream/sources/libs/nvdsinfer_customparser/libnvds_infercustomparser.so

[class-attrs-all]
nms-iou-threshold=0.5
pre-cluster-threshold=0.5
eps=0.2

SGIE0 labels.txt:
belt_helmet;right_helmet;wrong_helmet

[mchi]

outputCoverageBuffer is data from the output layer , it should indicates the model generate wrong result.

[17396156679]

thanks,I have checked that it is caused by net-scale-factor error value, I have two new questions.

1. secondary gie1 depend on secondary gie0’s output result in my config file,but i found secondary gie1 is invalid and found a topic on this issue in the forum, can you provide some good suggestions? Secondary gie depend on another secondary gie - Intelligent Video Analytics / DeepStream SDK - NVIDIA Developer Forums

2.For secondary gie0’s output result,I found that the probability value of class-id 0 is always very small and inconsistent with the results I got during the training model process. why is there such a result?

[mchi]

Sorry! Missed this update. Do you still need the support about this?

BTW, please create new topic for new issue