• Hardware Platform (Jetson / GPU): rtx 3090
• DeepStream Version: 5.1
• TensorRT Version: 7.2.3
• NVIDIA GPU Driver Version (valid for GPU only): 460.73.01
• Issue Type( questions, new requirements, bugs): questions
I test deepstream-image-meta-test example from sample_apps to encode cropped objects from detection. If I understand well, the responsible function for the encoding of cropped objects is nvds_obj_enc_process function.
When I run deepstream-image-meta-test exmaple sample with the encoding of cropped objects, the hardware encoder is not used. Output from nvidia-smi dmon during the running sample:
gpu pwr gtemp mtemp sm mem enc dec mclk pclk
0 270 70 - 99 16 0 3 9501 1935
0 273 70 - 99 15 0 2 9501 1920
0 272 70 - 92 15 0 2 9501 1282
0 264 70 - 99 15 0 2 9501 1282
0 271 70 - 99 15 0 2 9501 1920
0 272 71 - 99 16 0 2 9501 1920
0 272 70 - 99 16 0 2 9501 1920
0 270 70 - 99 15 0 2 9501 1282
0 273 71 - 99 15 0 2 9501 1920
0 277 71 - 99 17 0 3 9501 1282
0 271 70 - 95 16 0 2 9501 1282
0 262 70 - 99 14 0 2 9501 1282
0 260 70 - 100 13 0 2 9501 1920
0 269 70 - 99 14 0 2 9501 1920
0 271 71 - 100 16 0 3 9501 1920
0 267 70 - 100 14 0 2 9501 1920
Why doesn’t nvds_obj_enc_process function use the HW encoder?