Jetson Nano YoloV3 performance

Robotics & Edge Computing Jetson & Embedded Systems Jetson Nano
1

Hi all,

I followed the instructions in the link below and tried trt-yolo-app for YoloV3 implementation. I was expecting the performance can be improved a lot comparing with my implementation on Intel NC stick 2. But the FP16 result is 572ms per frame as oppose to the 600ms per frame on Pi3 + Intel NC Stick 2 which has no significant difference. The testing samples, weights and config are all the same. For YoloV3-Tiny the Jetson Nano shows very impressive of 25 frame/sec over the 11 frame/sec on NCS2. How come the performance on YoloV3 not quite comparable?

Plus, I tried to config as INT8 precision. “Platform doesn’t support this precision” err message popped out before it was aborted.

[url]https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/blob/master/yolo/README.md[/url]

Thanks,
Zhe

2

Hi,

Nano doesn’t support INT8 operation.

Could you monitor the system status at the same time and share with us?
If the model takes too much memory, TensorRT may not be able to choose a best algorithm for it.

We have some profiling result here for your reference:
[url]https://devblogs.nvidia.com/jetson-nano-ai-computing/[/url]

Thanks.

3

Hi,

Here is the tegrastats logging during the whole inference phase, the engine file sized 176.74MB:

RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [16%@102,8%@102,4%@102,10%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@29C CPU@36.5C PMIC@100C GPU@31C AO@43.5C thermal@31.5C POM_5V_IN 7514/7514 POM_5V_GPU 4704/4704 POM_5V_CPU 150/150
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [21%@1428,17%@1428,4%@1428,13%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@29C CPU@34.5C PMIC@100C GPU@31C AO@40C thermal@33.75C POM_5V_IN 4652/6083 POM_5V_GPU 1655/3179 POM_5V_CPU 433/291
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [18%@102,17%@102,7%@102,5%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@29C CPU@36C PMIC@100C GPU@31.5C AO@43C thermal@32.25C POM_5V_IN 7489/6551 POM_5V_GPU 4259/3539 POM_5V_CPU 226/269
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [15%@102,11%@204,8%@204,13%@204] EMC_FREQ 0% GR3D_FREQ 99% PLL@29.5C CPU@36.5C PMIC@100C GPU@32C AO@44C thermal@34C POM_5V_IN 7689/6836 POM_5V_GPU 4892/3877 POM_5V_CPU 150/239
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [14%@102,21%@102,3%@102,19%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@29.5C CPU@36C PMIC@100C GPU@31.5C AO@43.5C thermal@34C POM_5V_IN 6472/6763 POM_5V_GPU 3461/3794 POM_5V_CPU 346/261
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [23%@102,8%@102,10%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@30C CPU@36.5C PMIC@100C GPU@32C AO@44C thermal@33.5C POM_5V_IN 7640/6909 POM_5V_GPU 4720/3948 POM_5V_CPU 151/242
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [11%@614,8%@614,20%@614,5%@518] EMC_FREQ 0% GR3D_FREQ 0% PLL@29.5C CPU@34.5C PMIC@100C GPU@30.5C AO@40C thermal@34.25C POM_5V_IN 3064/6360 POM_5V_GPU 523/3459 POM_5V_CPU 281/248
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [13%@1428,22%@1428,9%@1428,7%@1428] EMC_FREQ 0% GR3D_FREQ 99% PLL@29.5C CPU@36C PMIC@100C GPU@31C AO@42.5C thermal@32C POM_5V_IN 5642/6270 POM_5V_GPU 2563/3347 POM_5V_CPU 583/290
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [10%@102,8%@102,15%@102,10%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@30C CPU@37C PMIC@100C GPU@32.5C AO@44.5C thermal@34.5C POM_5V_IN 7539/6411 POM_5V_GPU 4728/3500 POM_5V_CPU 151/274
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [18%@1428,11%@1428,9%@1428,8%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@30C CPU@35C PMIC@100C GPU@30.5C AO@40.5C thermal@34C POM_5V_IN 3987/6168 POM_5V_GPU 1074/3257 POM_5V_CPU 598/306
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [26%@102,3%@102,7%@102,8%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@30C CPU@36.5C PMIC@100C GPU@32C AO@44.5C thermal@34.5C POM_5V_IN 7678/6306 POM_5V_GPU 4311/3353 POM_5V_CPU 151/292
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [19%@102,15%@102,7%@102,12%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@30C CPU@37C PMIC@100C GPU@32.5C AO@43C thermal@34.5C POM_5V_IN 7929/6441 POM_5V_GPU 4546/3453 POM_5V_CPU 152/281
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [26%@307,6%@307,13%@307,13%@307] EMC_FREQ 0% GR3D_FREQ 99% PLL@30C CPU@36.5C PMIC@100C GPU@32C AO@43.5C thermal@33.75C POM_5V_IN 7132/6494 POM_5V_GPU 3998/3494 POM_5V_CPU 266/279
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [13%@102,15%@102,9%@102,6%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@30.5C CPU@37.5C PMIC@100C GPU@33C AO@45C thermal@35C POM_5V_IN 7640/6576 POM_5V_GPU 4917/3596 POM_5V_CPU 151/270
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [25%@518,5%@518,8%@518,14%@518] EMC_FREQ 0% GR3D_FREQ 99% PLL@30C CPU@36.5C PMIC@100C GPU@32.5C AO@43.5C thermal@34.5C POM_5V_IN 6654/6581 POM_5V_GPU 3603/3596 POM_5V_CPU 307/273
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [16%@102,17%@102,6%@102,7%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@31C CPU@38C PMIC@100C GPU@33C AO@45.5C thermal@34.5C POM_5V_IN 7615/6646 POM_5V_GPU 4765/3669 POM_5V_CPU 189/267
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [13%@307,16%@307,5%@307,20%@307] EMC_FREQ 0% GR3D_FREQ 55% PLL@30.5C CPU@36C PMIC@100C GPU@31.5C AO@42.5C thermal@35.25C POM_5V_IN 4699/6531 POM_5V_GPU 1613/3548 POM_5V_CPU 706/293
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [10%@102,4%@102,15%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@31C CPU@38C PMIC@100C GPU@33.5C AO@45.5C thermal@35C POM_5V_IN 7552/6588 POM_5V_GPU 4728/3614 POM_5V_CPU 151/285
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [19%@1428,16%@1428,4%@1428,10%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@31C CPU@36.5C PMIC@100C GPU@31.5C AO@41.5C thermal@34.5C POM_5V_IN 4173/6461 POM_5V_GPU 1232/3489 POM_5V_CPU 596/302
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [13%@102,14%@102,8%@102,10%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@31C CPU@37.5C PMIC@100C GPU@33.5C AO@45C thermal@35.5C POM_5V_IN 7729/6524 POM_5V_GPU 4349/3532 POM_5V_CPU 151/294
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [17%@102,14%@102,7%@102,8%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@31C CPU@38C PMIC@100C GPU@33C AO@43.5C thermal@35.25C POM_5V_IN 7003/6547 POM_5V_GPU 3840/3546 POM_5V_CPU 192/289
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [11%@102,19%@102,4%@102,7%@102] EMC_FREQ 0% GR3D_FREQ 0% PLL@30.5C CPU@35C PMIC@100C GPU@31.5C AO@40.5C thermal@36.25C POM_5V_IN 2434/6360 POM_5V_GPU 162/3392 POM_5V_CPU 162/283
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [18%@102,10%@102,9%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@31C CPU@37.5C PMIC@100C GPU@33C AO@45C thermal@34.25C POM_5V_IN 7640/6415 POM_5V_GPU 4417/3437 POM_5V_CPU 188/279
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [10%@307,16%@307,7%@307,9%@307] EMC_FREQ 0% GR3D_FREQ 99% PLL@31C CPU@38.5C PMIC@100C GPU@33.5C AO@45.5C thermal@36C POM_5V_IN 7615/6465 POM_5V_GPU 4682/3489 POM_5V_CPU 226/277
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [19%@102,10%@102,13%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@31C CPU@38C PMIC@100C GPU@33.5C AO@45C thermal@35.5C POM_5V_IN 7171/6494 POM_5V_GPU 4112/3514 POM_5V_CPU 228/275
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [25%@102,8%@102,5%@102,8%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@31.5C CPU@39C PMIC@100C GPU@34C AO@47C thermal@35.5C POM_5V_IN 7653/6538 POM_5V_GPU 4992/3571 POM_5V_CPU 151/270
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [22%@518,14%@518,7%@518,9%@518] EMC_FREQ 0% GR3D_FREQ 99% PLL@31C CPU@38C PMIC@100C GPU@33.5C AO@45C thermal@36.75C POM_5V_IN 6223/6526 POM_5V_GPU 3241/3558 POM_5V_CPU 423/276
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [12%@102,18%@102,6%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@31.5C CPU@38.5C PMIC@100C GPU@34.5C AO@46.5C thermal@36C POM_5V_IN 7653/6567 POM_5V_GPU 4690/3599 POM_5V_CPU 151/271
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [31%@1036,9%@1036,6%@1036,9%@1036] EMC_FREQ 0% GR3D_FREQ 4% PLL@31C CPU@36.5C PMIC@100C GPU@32C AO@42C thermal@35C POM_5V_IN 3410/6458 POM_5V_GPU 642/3497 POM_5V_CPU 521/280
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [23%@102,4%@102,6%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@32C CPU@38C PMIC@100C GPU@34C AO@46C thermal@36.25C POM_5V_IN 7628/6497 POM_5V_GPU 4395/3527 POM_5V_CPU 189/277
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [19%@1428,13%@1428,7%@1428,8%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@32C CPU@39C PMIC@100C GPU@33.5C AO@43.5C thermal@35.5C POM_5V_IN 5582/6467 POM_5V_GPU 2615/3497 POM_5V_CPU 312/278
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [30%@102,4%@102,4%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@32C CPU@38.5C PMIC@100C GPU@34C AO@46C thermal@36.75C POM_5V_IN 7640/6504 POM_5V_GPU 4493/3528 POM_5V_CPU 189/275
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [22%@102,5%@102,4%@102,12%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@32C CPU@39C PMIC@100C GPU@34.5C AO@47C thermal@36.5C POM_5V_IN 7615/6538 POM_5V_GPU 4773/3566 POM_5V_CPU 151/271
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [19%@204,11%@204,13%@204,8%@204] EMC_FREQ 0% GR3D_FREQ 99% PLL@32C CPU@39C PMIC@100C GPU@34.5C AO@45.5C thermal@35.5C POM_5V_IN 7042/6552 POM_5V_GPU 4005/3579 POM_5V_CPU 228/270
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [17%@204,17%@204,5%@204,6%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@32.5C CPU@39C PMIC@100C GPU@34.5C AO@47.5C thermal@36.75C POM_5V_IN 7628/6583 POM_5V_GPU 4879/3616 POM_5V_CPU 189/268
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [11%@614,16%@614,10%@614,14%@614] EMC_FREQ 0% GR3D_FREQ 99% PLL@32C CPU@37.5C PMIC@100C GPU@34C AO@45C thermal@35.75C POM_5V_IN 6060/6569 POM_5V_GPU 2947/3597 POM_5V_CPU 426/272
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [9%@102,12%@102,14%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@32.5C CPU@39.5C PMIC@100C GPU@35C AO@47.5C thermal@35.75C POM_5V_IN 7653/6598 POM_5V_GPU 4728/3628 POM_5V_CPU 151/269
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [10%@102,12%@102,15%@102,6%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@32C CPU@39C PMIC@100C GPU@34.5C AO@46.5C thermal@37.25C POM_5V_IN 7577/6624 POM_5V_GPU 4690/3656 POM_5V_CPU 189/267
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [15%@825,17%@825,8%@825,9%@825] EMC_FREQ 0% GR3D_FREQ 0% PLL@32C CPU@37.5C PMIC@100C GPU@33C AO@42.5C thermal@37.25C POM_5V_IN 3410/6541 POM_5V_GPU 682/3580 POM_5V_CPU 481/272
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [8%@102,10%@102,2%@102,17%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@32.5C CPU@39C PMIC@100C GPU@35C AO@47C thermal@37C POM_5V_IN 7653/6569 POM_5V_GPU 4538/3604 POM_5V_CPU 151/269
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [12%@1428,9%@1428,21%@1428,9%@1428] EMC_FREQ 0% GR3D_FREQ 20% PLL@33C CPU@39.5C PMIC@100C GPU@35C AO@44C thermal@35.25C POM_5V_IN 5875/6552 POM_5V_GPU 2801/3584 POM_5V_CPU 311/270
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [8%@102,13%@102,18%@102,8%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33C CPU@39.5C PMIC@100C GPU@35C AO@46.5C thermal@37.5C POM_5V_IN 7678/6579 POM_5V_GPU 4531/3607 POM_5V_CPU 189/268
RAM 2660/3965MB (lfb 160x4MB) SWAP 24/1024MB (cached 8MB) CPU [11%@102,12%@102,16%@102,7%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33C CPU@40C PMIC@100C GPU@35.5C AO@47.5C thermal@37C POM_5V_IN 7615/6603 POM_5V_GPU 4803/3634 POM_5V_CPU 151/266
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [10%@204,8%@204,25%@204,9%@204] EMC_FREQ 0% GR3D_FREQ 99% PLL@33C CPU@39.5C PMIC@100C GPU@35C AO@46.5C thermal@37.75C POM_5V_IN 7209/6617 POM_5V_GPU 4036/3643 POM_5V_CPU 342/267
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [13%@102,19%@102,9%@102,12%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33C CPU@40.5C PMIC@100C GPU@35.5C AO@48C thermal@37.5C POM_5V_IN 7653/6640 POM_5V_GPU 4954/3673 POM_5V_CPU 189/265
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [8%@710,17%@710,17%@710,11%@710] EMC_FREQ 0% GR3D_FREQ 99% PLL@33C CPU@39.5C PMIC@100C GPU@34.5C AO@46C thermal@35.75C POM_5V_IN 6021/6626 POM_5V_GPU 2870/3655 POM_5V_CPU 580/272
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [12%@102,9%@102,15%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33C CPU@40C PMIC@100C GPU@35.5C AO@47.5C thermal@35.75C POM_5V_IN 7590/6647 POM_5V_GPU 4765/3679 POM_5V_CPU 151/270
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [11%@102,16%@102,17%@102,9%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@33.5C CPU@38C PMIC@100C GPU@33.5C AO@43.5C thermal@38C POM_5V_IN 3565/6583 POM_5V_GPU 760/3618 POM_5V_CPU 600/277
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [11%@102,8%@102,16%@102,10%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33.5C CPU@40C PMIC@100C GPU@35.5C AO@47.5C thermal@37.25C POM_5V_IN 7653/6604 POM_5V_GPU 4432/3635 POM_5V_CPU 227/276
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [8%@1428,18%@1428,11%@1428,7%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@33.5C CPU@40.5C PMIC@100C GPU@35C AO@45C thermal@38C POM_5V_IN 5836/6589 POM_5V_GPU 2879/3619 POM_5V_CPU 272/275
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [7%@102,17%@102,17%@102,8%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33.5C CPU@40.5C PMIC@100C GPU@35.5C AO@47.5C thermal@38C POM_5V_IN 7653/6610 POM_5V_GPU 4425/3635 POM_5V_CPU 189/274
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [19%@102,10%@102,7%@102,8%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33C CPU@41C PMIC@100C GPU@36C AO@48.5C thermal@36.5C POM_5V_IN 7628/6629 POM_5V_GPU 4728/3656 POM_5V_CPU 226/273
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [12%@102,10%@102,10%@102,10%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33.5C CPU@41C PMIC@100C GPU@36C AO@47.5C thermal@37.25C POM_5V_IN 7916/6654 POM_5V_GPU 4862/3679 POM_5V_CPU 189/271
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [16%@102,6%@102,21%@102,5%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@33.5C CPU@40.5C PMIC@100C GPU@36C AO@48C thermal@38.25C POM_5V_IN 7590/6671 POM_5V_GPU 4387/3692 POM_5V_CPU 226/270
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [7%@102,7%@102,16%@102,11%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@34C CPU@41C PMIC@100C GPU@36C AO@48.5C thermal@38.25C POM_5V_IN 7753/6691 POM_5V_GPU 4908/3714 POM_5V_CPU 151/268
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [27%@403,11%@403,9%@403,7%@403] EMC_FREQ 0% GR3D_FREQ 99% PLL@33.5C CPU@40C PMIC@100C GPU@36C AO@47.5C thermal@36.5C POM_5V_IN 6703/6691 POM_5V_GPU 3615/3712 POM_5V_CPU 307/269
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [10%@204,17%@204,9%@204,11%@204] EMC_FREQ 0% GR3D_FREQ 99% PLL@33.5C CPU@41C PMIC@100C GPU@36.5C AO@48.5C thermal@38.75C POM_5V_IN 7628/6707 POM_5V_GPU 4773/3731 POM_5V_CPU 227/268
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [22%@1428,16%@1428,10%@1428,8%@1428] EMC_FREQ 0% GR3D_FREQ 15% PLL@33.5C CPU@39.5C PMIC@100C GPU@34.5C AO@45.5C thermal@38.75C POM_5V_IN 4173/6664 POM_5V_GPU 1071/3685 POM_5V_CPU 872/279
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [10%@102,13%@102,3%@102,20%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@34C CPU@41C PMIC@100C GPU@36.5C AO@49C thermal@37.75C POM_5V_IN 7628/6680 POM_5V_GPU 4622/3701 POM_5V_CPU 189/277
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [10%@1428,15%@1428,14%@1428,10%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@34C CPU@39C PMIC@100C GPU@35C AO@45.5C thermal@38.75C POM_5V_IN 5116/6654 POM_5V_GPU 1964/3672 POM_5V_CPU 472/280
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [14%@102,18%@102,5%@102,7%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@34C CPU@41C PMIC@100C GPU@36.5C AO@48.5C thermal@38.75C POM_5V_IN 7791/6673 POM_5V_GPU 4501/3686 POM_5V_CPU 189/279
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [8%@102,13%@102,14%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@34.5C CPU@41.5C PMIC@100C GPU@37C AO@49C thermal@37.25C POM_5V_IN 7603/6688 POM_5V_GPU 4660/3701 POM_5V_CPU 189/277
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [8%@204,9%@204,17%@204,19%@204] EMC_FREQ 0% GR3D_FREQ 99% PLL@34C CPU@40.5C PMIC@100C GPU@36.5C AO@48C thermal@38.75C POM_5V_IN 7042/6693 POM_5V_GPU 4005/3706 POM_5V_CPU 267/277
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [10%@102,5%@102,15%@102,10%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@34.5C CPU@42C PMIC@100C GPU@37C AO@50C thermal@38.25C POM_5V_IN 7691/6709 POM_5V_GPU 5030/3727 POM_5V_CPU 151/275
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [23%@307,14%@307,14%@307,9%@307] EMC_FREQ 0% GR3D_FREQ 99% PLL@34.5C CPU@40.5C PMIC@100C GPU@36.5C AO@48.5C thermal@39.25C POM_5V_IN 6233/6701 POM_5V_GPU 3247/3719 POM_5V_CPU 385/277
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [10%@102,3%@102,12%@102,13%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@35C CPU@41.5C PMIC@100C GPU@37C AO@49.5C thermal@39C POM_5V_IN 7628/6715 POM_5V_GPU 4736/3735 POM_5V_CPU 189/276
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [15%@403,13%@403,5%@403,8%@403] EMC_FREQ 0% GR3D_FREQ 19% PLL@34.5C CPU@41.5C PMIC@100C GPU@37C AO@46.5C thermal@39.25C POM_5V_IN 6311/6709 POM_5V_GPU 3296/3728 POM_5V_CPU 194/274
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [10%@403,22%@403,11%@403,6%@403] EMC_FREQ 0% GR3D_FREQ 99% PLL@34.5C CPU@41.5C PMIC@100C GPU@36.5C AO@48C thermal@39.25C POM_5V_IN 6703/6709 POM_5V_GPU 3615/3727 POM_5V_CPU 346/275
RAM 2660/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [8%@102,7%@102,14%@102,12%@102] EMC_FREQ 0% GR3D_FREQ 99% PLL@35C CPU@42C PMIC@100C GPU@37C AO@50C thermal@38.75C POM_5V_IN 7704/6724 POM_5V_GPU 4879/3743 POM_5V_CPU 189/274
RAM 2661/3965MB (lfb 160x4MB) SWAP 23/1024MB (cached 8MB) CPU [8%@1428,22%@1428,17%@1428,10%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@34.5C CPU@39.5C PMIC@100C GPU@35C AO@45.5C thermal@39.25C POM_5V_IN 3645/6680 POM_5V_GPU 480/3697 POM_5V_CPU 958/284
RAM 1991/3965MB (lfb 194x4MB) SWAP 23/1024MB (cached 8MB) CPU [21%@1428,14%@1428,13%@1428,21%@1428] EMC_FREQ 0% GR3D_FREQ 0% PLL@34.5C CPU@40C PMIC@100C GPU@35C AO@45C thermal@39.25C POM_5V_IN 3376/6633 POM_5V_GPU 200/3647 POM_5V_CPU 1085/295
RAM 1990/3965MB (lfb 194x4MB) SWAP 23/1024MB (cached 8MB) CPU [7%@102,5%@102,4%@102,8%@102] EMC_FREQ 0% GR3D_FREQ 0% PLL@34C CPU@39C PMIC@100C GPU@34C AO@44.5C thermal@36.75C POM_5V_IN 2119/6570 POM_5V_GPU 40/3597 POM_5V_CPU 122/293
RAM 1990/3965MB (lfb 194x4MB) SWAP 23/1024MB (cached 8MB) CPU [7%@102,3%@102,1%@102,11%@102] EMC_FREQ 0% GR3D_FREQ 0% PLL@34C CPU@38.5C PMIC@100C GPU@34C AO@44C thermal@36.25C POM_5V_IN 2116/6509 POM_5V_GPU 40/3549 POM_5V_CPU 162/291
RAM 1990/3965MB (lfb 194x4MB) SWAP 23/1024MB (cached 8MB) CPU [9%@102,5%@102,9%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 0% PLL@34C CPU@38.5C PMIC@100C GPU@34C AO@44C thermal@36.25C POM_5V_IN 2235/6452 POM_5V_GPU 81/3502 POM_5V_CPU 203/290
RAM 1990/3965MB (lfb 194x4MB) SWAP 23/1024MB (cached 8MB) CPU [8%@102,3%@102,1%@102,9%@102] EMC_FREQ 0% GR3D_FREQ 0% PLL@33.5C CPU@38.5C PMIC@100C GPU@34C AO@44C thermal@36.25C POM_5V_IN 2075/6393 POM_5V_GPU 40/3456 POM_5V_CPU 122/287
RAM 1990/3965MB (lfb 194x4MB) SWAP 23/1024MB (cached 8MB) CPU [23%@204,16%@204,23%@204,18%@204] EMC_FREQ 0% GR3D_FREQ 0% PLL@33.5C CPU@38.5C PMIC@100C GPU@34C AO@44.5C thermal@36.5C POM_5V_IN 1875/6334 POM_5V_GPU 122/3412 POM_5V_CPU 244/287

Here is the inference result:

Using previously generated plan file located at data/yolov3/yolov3-kHALF-kGPU-batch1.engine
Loading TRT Engine...
WARNING: Using an engine plan file across different models of devices is not recommended and is likely to affect performance or even cause errors.
Loading Complete!
Total number of images used for inference : 100
[======================================================================] 100 %
Network Type : yolov3 Precision : kHALF Batch Size : 1 Inference time per image : 573.856 ms

Batch Size = 4 : 521 ms
Batch Size = 8 : 517 ms

Thank you.

4

While I am doing the inference form testing images, I try to read form a PI camera and display the time when each frame is read, then the program crashes. Could YoloV3 be too heavy for Jetson Nano to handle?

5

Hi,

Could you help to check the root cause of the crash? Ex. dmesg.
Usually, crash is caused by memory issue rather than heavy loading.
It’s recommended to figure out the cause first.

By the way, batchsize means how many image you can feed into tensorRT one time.
So if you get T per inference time with batchsize b. The average time per image should be T/b.

Thanks.

6

Hi infohyu7s

Have you figured out the root cause of the crash?
Any result can be shared?

Thanks