dusty-nv / jetson-inference

Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.
https://developer.nvidia.com/embedded/twodaystoademo
MIT License
7.68k stars 2.96k forks source link

Distance of detection #1660

Open MorcretteApside opened 1 year ago

MorcretteApside commented 1 year ago

Hi,

I trained a detection model with detect net and with the pretrained model ssd mobilenet v1 on my jetson nano B01. I have 130 images of 1 drone split into train, test and validation. I'm running it on a camera IMX219 160IR. I successfully detect my drone but only in 50 cm range. Is there an easy way or not , to detect farther?

Thanks

dusty-nv commented 1 year ago

Hi @MorcretteApside, try running train_ssd.py with --resolution=512

Also, you will probably want to add more images to your dataset.

githubwalt commented 1 year ago

@MorcretteApside What is the resolution and aspect ratio of your images? Did Dusty's tip help?