Please provide complete information as applicable to your setup.
**• Hardware Platform (Jetson / GPU)Jetson nano
**• DeepStream Version6.0
**• JetPack Version (valid for Jetson only)4.6.1
**• TensorRT Version8.2
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs)
• How to reproduce the issue? (This is for bugs. Including which sample app is used, the configuration files content, the command line used, and other details for reproducing)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)
I intended to deploy retina face using Deepstream
the problem starts when I want to write a costume parser for it
the output of tensorrt engine(parser input) is in form of below:
output[0].shape=[6,4,0,0] (BBOX)
output[1].shape=[6,10,0,0] (landmarks)
output[2].shape=[6,2,0,0] (confidence)
for model, i used githup repo
for both downloading and converting to .onnx I used rensorrt command inside jetson to convert it to .engine model
for case of simplicity let’s discuss the Bbox output
when I run the pipeline tensorrt return a flattened bbox tensor which is 24 float. The Problem is that the values of these floats are so strange in each inference for example :
[-0.232422, -0.504525,-1.935834,0.5371099,…]
which to me has no meaning to parse(bbox coordinates)
remarkable is that when I change the height and width in the config file(from640480 to 1280960) the output is completely the same
I provide both app and pgie config so that we can correct any mistake or missing
[application]
[tiled-display]
enable=1
rows=1
columns=1
width=640
height=480
gpu-id=0
nvbuf-memory-type=0
[source0]
enable=1
type=3
uri=file://medium_0n69.mp4
num-sources=1
gpu-id=0
cudadec-memtype=0
[sink0]
enable=1
type=2
sync=0
gpu-id=0
nvbuf-memory-type=0
[osd]
enable=1
gpu-id=0
border-width=5
text-size=15
text-color=1;1;1;1;
text-bg-color=0.3;0.3;0.3;1
font=Serif
show-clock=0
clock-x-offset=800
clock-y-offset=820
clock-text-size=12
clock-color=1;0;0;0
nvbuf-memory-type=0
[streammux]
gpu-id=0
live-source=0
batch-size=1
batched-push-timeout=40000
width=640
height=480
enable-padding=0
nvbuf-memory-type=0
[primary-gie]
enable=1
gpu-id=0
gie-unique-id=1
nvbuf-memory-type=0
config-file=retina_config.txt
[tests]
file-loop=0
[property]
gpu-id=0
model-color-format=0
model-engine-file=/opt/models/retinaface/retina_r50.engine
labelfile-path=/opt/models/retinaface/labels.txt
process-mode=1
network-mode=2
gie-unique-id=1
network-type=0
output-blob-names=prob
maintain-aspect-ratio=1
batch-size=32
num-detected-classes=1
output-tensor-meta=1
parse-bbox-func-name=NvDsInferParseCustomRetinaface
custom-lib-path=///home/octoaiz/retinaface-v0/custom_parser/libnvdsinfer_our_custom_impl_retinaface.so
net-scale-factor=1.0
offsets=104.0;117.0;123.0
force-implicit-batch-dim=0
interval=0
scaling-compute-hw=1