Run Basler camera and Yolov8 Deep neural network on different platforms, Intel CPU, Intel UHD GPU, Nvidia GPU, Google coral USB and jetson nano.
This notebook includes different acquisition specs of the Camera including frame rate, brightness adjustment and ...
This notebook compares the performance of yolov8n on:
- Intel Core i7 CPU i7-8550U 1.80GHz
- Intel UHD 620 graphic GPU
- Nvidia GPU Geforce MX150 4G
Note: you should export the model to Openvino to get the best results for Intel processors.
Based on my results
Rank1: Yolov8n model with Nvidia GPU = 25.65 FPS
Rank2: Openvino model with Intel UHD 620 GPU = 11.98 FPS
Rank3: Yolov8n model with Google coral USB = 10.81
Rank4: Openvino model with CPU = 7.08 FPS
Rank5: Yolov8n model with CPU = 4.55 FPS
Rank6: Yolov8n model with jetson nano GPU = 4.3 FPS
Rank7: Openvino model with Nvidia GPU = 2.35 FPS
This Python script runs the Basler camera and Yolo model together and saves the frame if an object is detected
You can use the YoloCam script on Jetson Nano.
It is important to follow the steps in: https://i7y.org/en/yolov8-on-jetson-nano/.
You can only run this script on python3.8 on jetson nano and only in virtual environment using venv.
I could reach 4.3 FPS using Jetson nano CUDA.