access output from deepstream

Accelerated Computing Intelligent Video Analytics DeepStream SDK
#1

Hi,
I’m trying to run caffe model that gives out 4 tensors.
I need the output tensor to calculate the prob. of detection.
I’m trying to access the output tensors by
float* output_data1 = static_cast<float >(outputLayersInfo[0].buffer);
float output_data2 = static_cast<float >(outputLayersInfo[1].buffer);
float output_data3 = static_cast<float >(outputLayersInfo[2].buffer);
float output_data4 = static_cast<float *>(outputLayersInfo[3].buffer);

but I see that the results are much different than the results I get by running same picture in caffe.

  1. what can be the problem?
    2.is this is the right way to access the output tensors?

thanks

#2

Hi,

1. Please check if all the parameter in the config file is updated first.
You can file some information in our document:
>> Application Customization

2. You can check it in our source code here.
{deepstream_sdk_on_jetson}/sources

Thanks.

#3

I have used the example and still the results are far from what I expect to get
something that I noticed is when I change from BGR -> RGB the output doesn’t change

#4 
[property]
gpu-id=0
net-scale-factor=1
offsets=0;0;0;0
#0=RGB, 1=BGR
model-color-format=1
model-engine-file=V6.1.0_face_det.caffemodel_b1_fp32.engine
model-file=V6.1.0_face_det.caffemodel
proto-file=V6.1.0_face_det.prototxt
batch-size=1
## 0=FP32, 1=INT8, 2=FP16 mode
network-mode=0
num-detected-classes=21
interval=0
gie-unique-id=1
parse-func=0
is-classifier=0
output-blob-names=cls1;cls2;loc1;loc2
parse-bbox-func-name=NvDsInferParseCustomFasterRCNN
custom-lib-path=nvdsinfer_custom_impl_fasterRCNN/libnvdsinfer_custom_impl_fasterRCNN.so

[class-attrs-all]
threshold=0.2
eps=0.1
group-threshold=2
roi-top-offset=0
roi-bottom-offset=0
detected-min-w=0
detected-min-h=0
detected-max-w=0
detected-max-h=0

## Per class configuration
# Prevent background detection
[class-attrs-0]
threshold=1.1
#5 
[application]
enable-perf-measurement=1
perf-measurement-interval-sec=1
gie-kitti-output-dir=streamscl

[tiled-display]
enable=1
rows=1
columns=1
width=1920
height=1080

[source0]
enable=1
#Type - 1=CameraV4L2 2=URI 3=MultiURI
type=3
num-sources=1
uri=file://test10x2min_1.mp4

[streammux]
batch-size=1
batched-push-timeout=-1
## Set muxer output width and height
width=1920
height=1080

[sink0]
enable=1
#Type - 1=FakeSink 2=EglSink 3=File
type=2
sync=1
source-id=0

[osd]
enable=1
border-width=3
text-size=15
text-color=1;1;1;1;
text-bg-color=0.3;0.3;0.3;1
font=Arial
show-clock=0
clock-x-offset=800
clock-y-offset=820
clock-text-size=12
clock-color=1;0;0;0

[primary-gie]
enable=1
batch-size=1
gie-unique-id=1
interval=0
labelfile-path=labels.txt
model-engine-file=V6.1.0_face_det.caffemodel_b1_fp32.engine
config-file=config_infer_primary_fasterRCNN.txt
#6

how can I know if it even run my network?
I use the deepstream-app with my costume plugin

#7

Hi,

You can check the output log of deepstream.
By the way, we have a plugin sample for fasterRCNN. You can check it for some information first.
{deepstream_sdk_on_jetson}/sources/objectDetector_FasterRCNN

Thanks.

#8