Uploaded February 2020 | Updated September 2026, 2 weeks ago
Model was trained using 608x608 input.
you need to just change the resolution in cfg file, that's it.
YOLOv3 from darknet repo (fork by the famous AlexeyAB)
Code here: github.com/AlexeyAB/darknet
Model:
csresnext50-panet-spp-original-optimal.cfg - 65.4% mAP@0.5 (43.2% AP@0.5..0.95) - 35 FPS - 100.5 BFlops - 217 MB: csresnext50-panet-spp-original-optimal_final.weights
Model was trained using 608x608 input.
you need to just change the resolution in cfg file, that's it.
YOLOv3 from darknet repo (fork by the famous AlexeyAB)
Code here: github.com/AlexeyAB/darknet
Model:
csresnext50-panet-spp-original-optimal.cfg - 65.4% mAP@0.5 (43.2% AP@0.5..0.95) - 35 FPS - 100.5 BFlops - 217 MB: csresnext50-panet-spp-original-optimal_final.weights
![SSD Mobilenet v2 COCO TensorRT Jetson Nano
Network:
networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff
[TRT]
[TRT] Timing Report networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff
[TRT]
[TRT] Pre-Process CPU 0.05505ms CUDA 1.04162ms
[TRT] Network CPU 89.19215ms CUDA 45.69714ms
[TRT] Post-Process CPU 0.02917ms CUDA 0.02927ms
[TRT] Visualize CPU 0.23381ms CUDA 37.23661ms
[TRT] Total CPU 89.51019ms CUDA 84.00464ms
[TRT]
Run:
./detectnet output-codec=h265 overlay=box,labels,conf bitrate=50000000 headless log-level=info ~/0002-20170519-2.mp4 ~/ssd_mobilenet_v2_coco.mp4
Download my 4k video test sequence: https://archive.org/details/0002201705192 SSD Mobilenet v2 COCO TensorRT Jetson Nano](https://i.ytimg.com/vi/wPYgaA-Dm70/mqdefault.jpg)









