No labels in 3d/2d action recognition app

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 6.3 and 7.0(container)
• NVIDIA GPU Driver Version (valid for GPU only) 555
• Issue Type( questions, new requirements, bugs) bug
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
Run the deepstream 3d action recognition app with default values and 3d models

I am running the deepstream 3d action recognition app but getting no labels (only fps on osd),both 2d and 3d models with onnx as well as with etlt file gives same result, i.e. no labels. I am using pretrained models from ngc.
there are warning comings-
WARNING from element primary-nvinference-engine: nvinfer could not find input layer with name = input_rgb

Warning: nvinfer could not find input layer with name = input_rgb

0:00:32.819271434 3655 0x5577cfc6a6a0 WARN nvinfer gstnvinfer.cpp:1993:gst_nvinfer_process_tensor_input: warning: nvinfer could not find input layer with name = input_rgb

config_infer_primary_2d_action.txt

[property]
gpu-id=0

tlt-encoded-model=/media/sameer/Extras/deepstream-6.3/sources/apps/sample_apps/deepstream-3d-action-recognition_new/resnet18_2d_of_hmdb5_32_a100.etlt
tlt-model-key=nvidia_tao
model-engine-file=./resnet18_2d_rgb_hmdb5_32.etlt_b4_gpu0_fp16.engine

labelfile-path=labels.txt
batch-size=4
process-mode=1

requires preprocess metadata input

input-tensor-from-meta=1

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

network-mode=2
gie-unique-id=1

1: classifier, 100: custom

network-type=1

Let application to parse the inference tensor output

output-tensor-meta=1
tensor-meta-pool-size=8

config_preprocess_2d_custom.txt

[property]
enable=1
target-unique-ids=1

# 0=process on objects 1=process on frames

process-on-frame=1

# network-input-shape: batch, channel x sequence, height, width

2D sequence of 64 images

network-input-shape= 4;192;224;224

2D sequence of 32 images

network-input-shape= 4;96;224;224

# 0=RGB, 1=BGR, 2=GRAY

network-color-format=0
# 0=NCHW, 1=NHWC, 2=CUSTOM
network-input-order=2
# 0=FP32, 1=UINT8, 2=INT8, 3=UINT32, 4=INT32, 5=FP16
tensor-data-type=0
tensor-name=input_rgb

processing-width=224
processing-height=224

# 0=NVBUF_MEM_DEFAULT 1=NVBUF_MEM_CUDA_PINNED 2=NVBUF_MEM_CUDA_DEVICE
# 3=NVBUF_MEM_CUDA_UNIFIED  4=NVBUF_MEM_SURFACE_ARRAY(Jetson)

scaling-pool-memory-type=0
# 0=NvBufSurfTransformCompute_Default 1=NvBufSurfTransformCompute_GPU
# 2=NvBufSurfTransformCompute_VIC(Jetson)
scaling-pool-compute-hw=0
# Scaling Interpolation method
# 0=NvBufSurfTransformInter_Nearest 1=NvBufSurfTransformInter_Bilinear 2=NvBufSurfTransformInter_Algo1
# 3=NvBufSurfTransformInter_Algo2 4=NvBufSurfTransformInter_Algo3 5=NvBufSurfTransformInter_Algo4
# 6=NvBufSurfTransformInter_Default
scaling-filter=0

# model input tensor pool size

tensor-buf-pool-size=8

custom-lib-path=/opt/nvidia/deepstream/deepstream/lib/libnvds_custom_sequence_preprocess.so
#custom-lib-path=./custom_sequence_preprocess/libnvds_custom_sequence_preprocess.so
custom-tensor-preparation-function=CustomSequenceTensorPreparation

2D conv custom params

[user-configs]
channel-scale-factors=0.007843137;0.007843137;0.007843137
channel-mean-offsets=127.5;127.5;127.5
stride=1
subsample=0

[group-0]
src-ids=0;1;2;3
process-on-roi=1
roi-params-src-0=0;0;1280;720
roi-params-src-1=0;0;1280;720
roi-params-src-2=0;0;1280;720
roi-params-src-3=0;0;1280;720

Are you using your own video source? Could you attach that? Thanks

all sample videos only, nothing changed in configs

Is your model the latest version?

yes, downloaded from https://ngc.nvidia.com/catalog/models/nvidia:tao:actionrecognitionnet , tried both onnx and etlt as well as 2d / 3d models , no labels just fps on screen

OK. You need to modify the config file according to the model you are actually using.
You can check the input layer name of the model first, then modify the tensor-name=input_rgb in the config file.

how can i check input layer name of model , model is pretrained

If you are using the onnx model, you can check that with any onnx parse tool, like netron.
If you are using etlt model, you can run that with DeepStream first. The detailed layer information is in the log file.
Also, DeepStream can only support the RGB model, not optical flow model.

issue is solved

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.