I have downloaded DeepStrean 1.5 and compile all given examples. All examples are compiled fine in the Ubuntu 16.04 LTS on a P4 server but when I try to run the NVDECINFER_DETECTION example, it fails after reading the model’s description:
~/deepstream/samples/data/video ~/deepstream/samples/nvDecInfer_detection
~/deepstream/samples/nvDecInfer_detection
[DEBUG][15:43:23] Device ID for display [0]: Tesla P4
[DEBUG][15:43:23] Device ID for inference [0]: Tesla P4
[DEBUG][15:43:23] Video channels: 4
[ERROR][15:43:23] Warning: No mean files.
[DEBUG][15:43:23] GUI enabled.
[DEBUG][15:43:23] Endless Loop: 0
[DEBUG][15:43:23] Device name: Tesla P4
[DEBUG][15:43:24] Use INT8 data type.
[DEBUG][15:43:28] =========== Network Parameters Begin ===========
[DEBUG][15:43:28] Network Input:
[DEBUG][15:43:28] >Batch :4
[DEBUG][15:43:28] >Channel :3
[DEBUG][15:43:28] >Height :368
[DEBUG][15:43:28] >Width :640
[DEBUG][15:43:28] Network Output [0]
[DEBUG][15:43:28] >Channel :4
[DEBUG][15:43:28] >Height :23
[DEBUG][15:43:28] >Width :40
[DEBUG][15:43:28] Network Output [1]
[DEBUG][15:43:28] >Channel :16
[DEBUG][15:43:28] >Height :23
[DEBUG][15:43:28] >Width :40
[DEBUG][15:43:28] =========== Network Parameters End ===========
[ERROR][15:43:28] CUDA error 304 at line 210 in file presenterGL.cpp
./run.sh: line 43: 6027 Segmentation fault (core dumped) ../bin/sample_detection -devID_display=${DISPLAY_GPU} -devID_infer=${INFER_GPU} -nChannels=${CHANNELS} -fileList=${FILE_LIST} -deployFile=${DEPLOY} -modelFile=${MODEL} -labelFile=${LABEL} -int8=1 -calibrationTableFile=${CALIBRATION} -tileWidth=${TILE_WIDTH} -tileHeight=${TILE_HEIGHT} -tilesInRow=${TILES_IN_ROW} -fullscreen=0 -gui=1 -endlessLoop=0
This error occurs when gui=1 is used, and there is no error when gui=0. The gui value can be change in the run.sh