Deepstream_Yolo-Face

[MadddY-1999]

i have setup all of the deepstream SDK with python bindings, and i have tested their deepstream-test1 app and its working absolutely fine, so now i wanted to work with yolo face model to run it in deepstream pipeline and i have successfully followed this instruction's here: https://github.com/marcoslucianops/DeepStream-Yolo-Face

and afterwards when i try to run the pipeline im getting this error here:

root@AAM-LAPTOP-027:/opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face# python3 deepstream.py -s file:///opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face/test.mp4 -c config_infer_primary_yoloV8_face.txt
/opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face/deepstream.py:201: DeprecationWarning: Gst.Element.get_request_pad is deprecated
streammux_sink_pad = streammux.get_request_pad(pad_name)

SOURCE: file:///opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face/test.mp4
CONFIG_INFER: config_infer_primary_yoloV8_face.txt
STREAMMUX_BATCH_SIZE: 1
STREAMMUX_WIDTH: 1920
STREAMMUX_HEIGHT: 1080
GPU_ID: 0
PERF_MEASUREMENT_INTERVAL_SEC: 5
JETSON: FALSE

gstnvtracker: Loading low-level lib at /opt/nvidia/deepstream/deepstream/lib/libnvds_nvmultiobjecttracker.so
[NvMultiObjectTracker] Initialized
0:00:01.075958955 3707 0x559ed1ea7c70 INFO nvinfer gstnvinfer.cpp:684:gst_nvinfer_logger: NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::deserializeEngineAndBackend() <nvdsinfer_context_impl.cpp:2092> [UID = 1]: deserialized trt engine from :/opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face/yolov8n-face.onnx_b1_gpu0_fp32.engine
Implicit layer support has been deprecated
INFO: ../nvdsinfer/nvdsinfer_model_builder.cpp:327 [Implicit Engine Info]: layers num: 0

0:00:01.076020472 3707 0x559ed1ea7c70 INFO nvinfer gstnvinfer.cpp:684:gst_nvinfer_logger: NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::generateBackendContext() <nvdsinfer_context_impl.cpp:2195> [UID = 1]: Use deserialized engine model: /opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face/yolov8n-face.onnx_b1_gpu0_fp32.engine
0:00:01.080324774 3707 0x559ed1ea7c70 INFO nvinfer gstnvinfer_impl.cpp:343:notifyLoadModelStatus: [UID 1]: Load new model:config_infer_primary_yoloV8_face.txt sucessfully

Failed to query video capabilities: Inappropriate ioctl for device
Segmentation fault (core dumped)

i have done everything but not able to resolve this issue here, i would like some expert here ot help me out here please.
i would be very grateful for the kindness.

for reference im using the deepstream sdk 7.1 and my cuda_ver = 12.6

[MadddY-1999]

here is the config file:
[property]
gpu-id=0
net-scale-factor=0.0039215697906911373
model-color-format=0
onnx-file=/opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face/yolov8n-face.onnx
model-engine-file=/opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face/yolov8n-face.onnx_b1_gpu0_fp32.engine
#int8-calib-file=calib.table
labelfile-path=/opt/nvidia/deepstream/deepstream-7.1/sources/DeepStream-Yolo-Face/labels.txt
batch-size=1
network-mode=0
num-detected-classes=1
interval=0
gie-unique-id=1
process-mode=1
network-type=3
cluster-mode=4
maintain-aspect-ratio=1
symmetric-padding=1
#workspace-size=6000
parse-bbox-instance-mask-func-name=NvDsInferParseYoloFace
custom-lib-path=nvdsinfer_custom_impl_Yolo_face/libnvdsinfer_custom_impl_Yolo_face.so
output-instance-mask=0

[class-attrs-all]
pre-cluster-threshold=0.25
topk=300

[fdarvas]

looks like no solution so far - if it helps , I have run into the exact same issue with a custom RT-DETR model, also using deepstream 7.1 on an x86 with an L4. This same model (i.e. the locally built for archticture engine) does work with DS 6.4 on a Orin with JetPack 6.0 .

[MadddY-1999]

the model conversion was not right, there was a file in utils to export the .pt model to onnx, use that one for conversion, then it will work.

[fdarvas]

I have a correct model - an RT-DETR model , which looks ok in netron. I.e. it has and input layer named "input" and three output tensors, i.e. boxes, scores and classes. But it fails in line 109 of nvdsparsebbox_Yolo.cpp const uint outputSize = output.inferDims.d[0]; ; Unclear why DS 7.1 would return an empty vector here?

[MadddY-1999]

this file here nvdsparsebbox_Yolo.cpp is created for the yolo face model, so you might have to create your own parsing file for it if you want to use your own model instead.