Gstnvinfer "can not find binding of given name"

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Jetson Xavier agx developer kit
jetpack 4.3
deepstream 4

I have created my own yolov3-tiny model and used /opt/nvidia/deepstream/deepstream-5.0/sources/objectDetector_Yolo to create a new libnvdsinfer_custom_impl_Yolo.so file

when I run my model I get eh follow errors:

[NvDCF] Initialized
Deserialize yoloLayerV3 plugin: yolo_17
Deserialize yoloLayerV3 plugin: yolo_24
0:00:05.745702759 e[336m16643e[00m 0x5589959690 e[31;01mERROR e[00m e[00m nvinfer gstnvinfer.cpp:511:gst_nvinfer_logger:<primary_gie_classifier>e[00m NvDsInferContext[UID 1]:log(): INVALID_ARGUMENT: Can not find binding of given name
0:00:05.745791563 e[336m16643e[00m 0x5589959690 e[33;01mWARN e[00m e[00m nvinfer gstnvinfer.cpp:515:gst_nvinfer_logger:<primary_gie_classifier>e[00m NvDsInferContext[UID 1]:checkEngineParams(): Could not find output layer ‘bboxes’ in engine
0:00:05.745839598 e[336m16643e[00m 0x5589959690 e[31;01mERROR e[00m e[00m nvinfer gstnvinfer.cpp:511:gst_nvinfer_logger:<primary_gie_classifier>e[00m NvDsInferContext[UID 1]:log(): INVALID_ARGUMENT: Can not find binding of given name
0:00:05.745880304 e[336m16643e[00m 0x5589959690 e[33;01mWARN e[00m e[00m nvinfer gstnvinfer.cpp:515:gst_nvinfer_logger:<primary_gie_classifier>e[00m NvDsInferContext[UID 1]:checkEngineParams(): Could not find output layer ‘coverage’ in engine
cb_sourcesetup set 100 latency

Summary:

gstnvinfer.cpp:511 - Can not find binding of given name
gstnvinfer.cpp:515 - Could not find output layer ‘bboxes’ in engine
gstnvinfer.cpp:515 - Could not find output layer ‘coverage’ in engine

Is there a problem with my config files or my model?

thanks Chris

3

Is your yolov3 model trained and generated according to https://docs.nvidia.com/metropolis/TLT/tlt-getting-started-guide/? If so, you can use /opt/nvidia/deepstream/deepstream-5.0/sources/objectDetector_Yolo, or else, it will fail.

4

Thanks I’ll check everything over again

The model works well but I wondered what the errors mean? And how I can get rid of them?

I’m using deepstream 4 not 5 does that make make a difference

5

For deepstream 5, TensorRT version is different. And I think your model’ output layers are not ‘bboxes’ and ‘coverage’

6

is there any documentation specifically for deepstream4 version?

the TLT 2.0 documentation has no mention of given name at all?

thanks

7

Is this because I had a yolo dataset and not kitti? so the txt file structure is different?

model engine files created correctly and the model appears to work

Is there a way to convert yolo .txt files to kitti format?

8

Deepstream 4.0 is not supported any longer. Please upgrade to latest deepstreamSDK. https://developer.nvidia.com/deepstream-getting-started

9

There is no access to the old documents?

I can’t upgrade yet, I must use DS4 for the time being

thanks

10

Do you mean document with DS 4.0? If so, you can refer to https://docs.nvidia.com/metropolis/deepstream/4.0.2/dev-guide/index.html#page/DeepStream%2520Development%2520Guide%2Fdeepstream_quick_start.html%23. But it will not help with your model issues.

11

no, I wondered if there was an older version of the TLT documentation suitable for use with DS4?

thanks

12

No. TLT2.0 support yolov3