yolov11n_raspberrypi5

YOLOv11n for Raspberry Pi 5: Efficient Object Detection

The YOLOv11n project provides a highly optimized object detection solution for the Raspberry Pi 5, utilizing advanced ARM NEON intrinsics. The core of this repository is a prebuilt library that facilitates quick and efficient inference of the YOLOv11n model. This implementation is designed not only for performance but also for ease of use in embedded systems.

Key Features

• Optimized Performance: The C implementation runs the complete 24-layer YOLOv11n model on images such as data/dog.jpg (768×576) in approximately 310 ms, which is about 1.4 times faster than the corresponding Python version while maintaining identical detection levels.
• No Dependencies Required: There is no need for additional installations, including Python or deep learning frameworks, ensuring a quick setup and execution.
• Minimal Start-Up Time: The compiled binary executes in milliseconds without the delays typically introduced by interpreters or package environments, providing instant object detection capabilities.
• Low Memory Usage: Operating without a Python runtime or heavy frameworks leads to a compact memory footprint, ideal for embedded and headless applications.
• Offline and Standalone Operation: The model runs entirely offline, making it suitable for environments where internet access is limited or unavailable.
• Single Prebuilt Library: The comprehensive neural network, its weights, and post-processing functions reside within lib/libyolov11n.so. A small demo command-line interface utilizes the public API found in inc/yolov11n.h.

Performance Metrics

The performance was evaluated on the Raspberry Pi 5 with a Cortex-A76 processor featuring four cores. Below are the latency results along with the number of detections for both implementations:

Both implementations produce consistent detections from the same input image:

C (NEON):                            Python (ultralytics):
  bicycle: 0.937 [128,136,567,419]    dog:     0.921 [132,221,310,540]
  dog:     0.930 [132,221,310,540]    bicycle: 0.914 [129,137,565,418]
  truck:   0.501 [469,74,691,173]     truck:   0.499 [470,75,692,171]
  car:     0.490 [471,74,692,172]     car:     0.458 [471,75,693,171]

This project targets users seeking an efficient and effective object detection solution for the Raspberry Pi environment, combining ease of use with high performance.

Package Structure

The repository is organized as follows:

yolov11n_raspberry_pi/
├── run.sh                    # Quick build & run script (C or --py, --bench)
├── CMakeLists.txt            # Builds the demo CLI against the prebuilt library
├── README.md                 # This file
├── LICENSE                   # MIT License
├── lib/
│   └── libyolov11n.so       # Prebuilt inference engine (ARM NEON, ~100-frame trial)
├── inc/
│   └── yolov11n.h           # Public API (only header needed to use the library)
├── src/
│   └── main.c               # Demo CLI using the public API
├── python/
│   ├── predict.py           # Official ultralytics reference
│   └── benchmark.py         # C vs Python benchmark
└── data/
    ├── dog.jpg / bus.jpg    # Sample images
    └── coco_names.txt       # COCO class names

This structured layout facilitates easy navigation and utilization of the library, while also ensuring that all necessary components for running and testing the object detection capabilities are readily available.