Here is my fall detection with jetson nano and raspberry camera module.
The project uses TRT_pose model with 224x224 resolution and manages around 10-15 fps. The fall detection checks if nose gets too close to hips or ankles and then changes the color of drawing the pose. More information on my github.
I am trying to run the trt_pose model on Jetson Nano 2GB model, but the application freezes due to low ram. How did you achieve 10-15 FPS? Did you face any issue while using TRT pose? I followed the exact installation instructions mentioned here - GitHub - NVIDIA-AI-IOT/trt_pose: Real-time pose estimation accelerated with NVIDIA TensorRT
Any help would be appreciated, Thank you