Confusion in writing Custom Config.txt for deeptsream 6.3 and hence the custom python app

[muhammad.fasih1]

**• Hardware Platform Ubuntu 20.04 with Nvidia GeForce GTX 1080 Ti
**• DeepStream 6.3-triton-multiarch Docker image
**• TensorRT 8.5.3
**• NVIDIA GPU Driver version 555.42.02
**• Issue Type: Aborted (core dumped)

**• Detailed Error:

Frames will be saved in  frames
Creating streamux 
 
Creating source_bin  0  
 
Creating source bin
source-bin-00
Creating Pgie 

Unknown or legacy key specified 'input-blob-names' for group [property]
Creating EGLSink 

Playing file file:///opt/nvidia/deepstream/deepstream-6.3/sources/deepstream_python_apps/apps/deepstream-tickerdetection/test.mp4
Adding elements to Pipeline 

Linking elements in the Pipeline 

Linking elements in the Pipeline 

Starting pipeline 

0:00:01.942882558 13033      0x2c3e090 INFO                 nvinfer gstnvinfer.cpp:682:gst_nvinfer_logger:<primary-inference> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::deserializeEngineAndBackend() <nvdsinfer_context_impl.cpp:1988> [UID = 1]: deserialized trt engine from :/opt/nvidia/deepstream/deepstream-6.3/sources/deepstream_python_apps/apps/deepstream-tickerdetection/det_model.engine
WARNING: [TRT]: The getMaxBatchSize() function should not be used with an engine built from a network created with NetworkDefinitionCreationFlag::kEXPLICIT_BATCH flag. This function will always return 1.
INFO: ../nvdsinfer/nvdsinfer_model_builder.cpp:610 [Implicit Engine Info]: layers num: 2
0   INPUT  kFLOAT input           360x640         
1   OUTPUT kFLOAT output          180x320         

python3: nvdsinfer_backend.cpp:135: virtual bool nvdsinfer::TrtBackendContext::canSupportBatchDims(int, const NvDsInferBatchDims&): Assertion `m_AllLayers[bindingIdx].inferDims.numDims == batchDims.dims.numDims' failed.
Aborted (core dumped)

I think, I am unable to write correct config.txt file. My current config.txt file is as follows:

[property]
gpu-id=0
net-scale-factor=0.0039215697906911373
tlt-model-key=none
#infer-dims=1;360;640
onnx-file=/opt/nvidia/deepstream/deepstream-6.3/sources/deepstream_python_apps/apps/deepstream-tickerdetection/det_model.onnx
model-engine-file=/opt/nvidia/deepstream/deepstream-6.3/sources/deepstream_python_apps/apps/deepstream-tickerdetection/det_model.engine
#labelfile-path=</path/to/your/label_file.txt>
#input-blob-names=input
output-blob-names=output
infer-dims=360;640;1
input-blob-names=input;info
force-implicit-batch-dim=1
batch-size=1
## 0=FP32, 1=INT8, 2=FP16 mode
network-mode=0
interval=0
gie-unique-id=1
#change the input to gray scale
model-color-format=1
network-type=3
#output-blob-names=output
#parse-bbox-func-name=NvDsInferParseCustomSegMask
#custom-lib-path=</path/to/your/customparser.so>

The model I’m using is a custom segmentation model that takes in gray scale images with dimensions, The model takes input as [1,360,640] and outputs [1, 180, 320] as the binary mask. Here when i run python app, it also says that

Unknown or legacy key specified 'input-blob-names' for group [property]

I have read that it’s important to specify the input and output layers names of the tensors being used in the model. Any help on this will be much appreciated.

[yuweiw]

You can refer to our onnx demo deepstream_yolo to write the config file. Please refer to our Guide to learn some basic meaning of the paramters too.

[yingliu]

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks