MegaDetector V6

The latest release focuses on efficiency, modern architectures, and licensing flexibility — smaller, faster, better.

Highlights

• 50x smaller — the compact YOLOv10 variant has 2.3M parameters (2% of V5's 139.9M) with comparable accuracy
• Multiple architectures — YOLOv9, YOLOv10, and RT-DETR variants; pick the one that fits your hardware
• Permissive licenses — MIT and Apache-2.0 options alongside AGPL-3.0
• Ongoing fine-tuning — V6 models are continuously fine-tuned on newly collected public and private data

Quick Start

pip install PytorchWildlife

from PytorchWildlife.models import detection as pw_detection
model = pw_detection.MegaDetectorV6()
results = model.single_image_detection("path/to/image.jpg")

Resources

• Full documentation
• Model Zoo
• Try it on Hugging Face
• Google Colab notebook

MegaDetector is part of the microsoft/Biodiversity ecosystem and invoked through PyTorch Wildlife.