YOLOv8 Object Detection on Jetson Nano | Real-Time AI Demonstration only for JETSON NANO

YOLOv8 Object Detection on Jetson Nano
Author:
Darshan Anand
Pre-final Year CSE-AIML Student
Dayananda Sagar University
Email: darshananand004@gmail.com
Installation Steps

    1. Install Jetpack 4.6 on Jetson Nano
      Ensure Jetpack 4.6 (L4T 32.6.1) is installed on the Jetson Nano.
    1. Update and Install Dependencies
      Open a terminal and run the following commands:
sudo apt update 
sudo apt install -y python3.8 python3.8-venv python3.8-dev python3-pip libopenmpi
dev libomp-dev libopenblas-dev libblas-dev libeigen3-dev libcublas-dev 
    1. Clone the YOLOv8 Repository
git clone https://github.com/ultralytics/ultralytics 
cd ultralytics 
    1. Create and Activate a Python 3.8 Virtual Environment
python3.8 -m venv venv 
source venv/bin/activate 
    1. Update Python Packages
pip install -U pip wheel gdown 
    1. Download and Install Pre-built PyTorch and TorchVision Packages
# PyTorch 1.11.0 
gdown https://drive.google.com/uc?id=1hs9HM0XJ2LPFghcn7ZMOs5qu5HexPXwM 
# TorchVision 0.12.0 
gdown https://drive.google.com/uc?id=1m0d8ruUY8RvCP9eVjZw4Nc8LAwM8yuGV 
python3.8 -m pip install torch-*.whl torchvision-*.whl 
    1. Install YOLOv8 Python Package
pip install . 
    1. Execute Object Detection
yolo task=detect mode=predict model=yolov8n.pt source=0 show=True 
yolo task=segment mode=predict model=yolov8n-seg.pt source=0 show=True 

Observations and Results
Object Detection Output 1: 480x640 1 person, 170.5ms …
[Image: file-JGLGw584df2vKiawc0BHVC2i not found]
Object Detection Output 2: 480x640 1 person, 165.0ms …
[Image: file-VyiMQGF0kqh2lJfEczALIXkn not found]
Object Detection Output 3: 480x640 1 person, 163.8ms …
[Image: file-JmUKhF7FzBkzRCSruKhajKwq not found]
Object Detection Output 4: 480x640 2 persons, 167.1ms …
[Image: file-bQ3NkNaiMFMHBHI1ucgdKW7G not found]
Performance Metrics
The YOLOv8 model achieved the following FPS (frames per second) on the Jetson Nano for
object detection and segmentation tasks:

Model Detect FPS Segment FPS
yolov8n 6.1
4.2
yolov8s 3.1
yolov8m 1.3
yolov8l 0.77
yolov8x 0.48
Conclusion
2.2
0.96
0.61
0.38
The project successfully demonstrated the implementation of YOLOv8 for real-time object
detection on the Jetson Nano. The setup and installation were straightforward, and the
model performed adequately given the hardware constraints of the Jetson Nano. Future
work could involve optimizing the model further for better performance or exploring more
complex detection tasks.
YOLOv8, which stands for You Only Look Once version 8, is the latest iteration of a series of
models known for their speed and accuracy in object detection tasks. It builds upon the
successes of its predecessors by incorporating more advanced techniques and
optimizations, making it a suitable choice for real-time applications on resource
constrained devices like the Jetson Nano.
One of the primary challenges encountered during this project was managing the limited
computational resources of the Jetson Nano. Despite this, YOLOv8 performed remarkably
well, achieving a respectable frame rate that would be sufficient for many real-time
applications. This speaks volumes about the efficiency of the YOLOv8 architecture.
Future improvements to this project could include further optimization of the YOLOv8
model for the Jetson Nano. This could involve techniques such as model pruning and
quantization to reduce the model size and increase inference speed. Additionally, exploring
other versions of YOLOv8 (like yolov8s, yolov8m, yolov8l, and yolov8x) could provide
insights into the trade-offs between speed and accuracy.
Another potential area of exploration is integrating YOLOv8 with other sensors and systems
on the Jetson Nano. For instance, combining YOLOv8 with depth sensors could enhance
object detection capabilities by providing additional context about the environment.
Similarly, integrating with robotic systems could enable autonomous navigation and
interaction based on real-time object detection.
Overall, this project demonstrates the viability of deploying advanced AI models on edge
devices, opening up numerous possibilities for real-time, on-device intelligence in various
applications.
Contact
Darshan Anand
Email: darshananand004@gmail.com
Dayananda Sagar University
CSE - AIML Pre-final Year Student