Object Detection -- YOLOv7

This package performs object detection using the YOLOv7 model with ROS 2. It subscribes to camera image data and publishes detection results on a specified ROS 2 topic. The node can be used for real-time object detection in a ROS 2 environment.

Prerequisites

• ROS 2 Humble: Ensure that your environment is set up with ROS 2 Humble.
• PyTorch: Deep learning framework required for YOLOv7. Install the appropriate version based on your environment (e.g., CPU or CUDA).
• Python Packages: Additional Python packages needed for various functionalities:
  • tqdm
  • pandas
  • requests
  • seaborn

This node has been tested with the following software configuration:

• Operating System: Ubuntu 22.04

• Python: 3.10.12

• ROS 2: Humble

• CUDA: 11.8

• Python packages::

  • NumPy: 1.24.4
  • OpenCV-Python: 4.9.0.80
  • PyTorch: 2.2.2+cu118
  • Torchvision: 0.17.2+cu118
  • Torchaudio: 2.2.2+cu118
  • tqdm: 4.66.2
  • pandas: 2.2.2
  • requests: 2.31.0
  • seaborn: 0.13.2

Usage

Step 1: Install the ROS 2 Package

• Clone the Git repository to your workspace

$ git clone https://github.com/AndersonYu7/YOLOv7_ROS2.git detect

Step 2: Ensure the Camera is Running

• Make sure your camera is running
• Ensure it's publishing image data to the /image/image_raw topic.

Step 3: Launch the detect program

$ ros2 launch yolov7_obj_detect object_detection_launch.py

Detection results

This node publishes detection results on the /detect/objs topic. Each message contains the following information:

• Labels: The names of the detected objects.
• Scores: The confidence scores for each detection.
• Bounding Boxes: The coordinates of the bounding boxes for each detected object, represented as (xmin, ymin, xmax, ymax).

Parameter Description

• weights:

  • Provide the name of the YOLOv7 model weights file

• conf_thres: Confidence threshold for object detection.

  • Set the confidence threshold (conf_thres) to a value between 0 and 1.

• iou_thres: Intersection over Union (IOU) threshold for object detection.

  • Set the IOU threshold (iou_thres) to a value between 0 and 1.

• device: Specify the device for object detection. This parameter allows inputs like "cpu" or "0,1,2,3". By default, the object detection algorithm uses CUDA on GPUs if available, otherwise it uses CPU.

  • Provide the device (device) where the object detection algorithm will be executed. Use "cpu" for CPU or specify GPU device IDs like "0,1,2,3".

• img_size: Image size for object detection.

  • Set the image size (img_size) to a suitable value, such as 640, depending on the requirements of the object detection model.

• show_img:

  • Whether to display the detection results.

Model File Location

The YOLOv7 model weights file should be placed in the following location:

• detect/yolov7_obj_detect/weights: Store the YOLOv7 model weights file in this directory.

Notes

Ensure that your camera is properly configured and running. If you're not receiving images, check the camera connection and topic name.

References

• YOLOv7
• YOLOv7 ROS2