Why numbers of anchor are 3 in yolov5-p2.yaml?

[DiaJB]

Search before asking

• I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

There are 4 detectors([21, 24, 27, 30]) in in yolov5-p2.yaml. Why numbers of anchor are 3 in yolov5-p2.yaml? Did anyone try it and whether it worked?

Additional

No response

[UltralyticsAssistant]

👋 Hello @DiaJB, thank you for your interest in YOLOv5 🚀! This is an automated response to assist you, and an Ultralytics engineer will also review and address your question soon.

Regarding your question about the number of anchors in yolov5-p2.yaml, if this is a 🐛 Bug Report, please provide a minimum reproducible example to help us better understand and address the issue.

For general inquiries, please ensure you are following our training guidelines and share any relevant details such as your dataset examples and training logs to assist in providing an accurate response.

Requirements

Ensure you have [Python>=3.8.0] installed with all necessary dependencies, including [PyTorch>=1.8]. If you haven't already, you'll need to clone the yolov5 repository and install the requirements to get started:

git clone <repository_url>  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 can run on various platforms with preinstalled dependencies, including online notebooks with free GPU access, Google Cloud, Amazon Web Services, and Docker Containers.

Status

The continuous integration (CI) tests for YOLOv5 are scheduled to run frequently, ensuring correct operation of all functionalities including training, validation, inference, export, and benchmarks across different operating systems.

Your input and active participation are invaluable, and we appreciate your contribution to the YOLOv5 community. Feel free to provide any additional information or clarifications needed. 🤖✨

[pderrenger]

The use of 3 anchors per detector in yolov5-p2.yaml is a design choice to balance model complexity and performance, as three anchors typically provide sufficient flexibility for most object detection tasks. If you have specific needs, you can experiment with different anchor configurations to see what works best for your dataset.