YOLOv12

YOLOv12 is a newly proposed attention-centric variant of the YOLO family that focuses on incorporating efficient attention mechanisms into the backbone while preserving real-time performance. Instead of relying heavily on CNN-based architectures like its predecessors, YOLOv12 introduces a simple yet powerful “area attention” module, which strategically partitions the feature map to reduce the quadratic complexity of full self-attention. It also adopts residual efficient layer aggregation networks (R-ELAN) to enhance feature aggregation and training stability, especially for larger models. These innovations, together with refinements such as scaled residual connections and a reduced MLP ratio, enable YOLOv12 to harness the benefits of attention (e.g., better global context modeling) without sacrificing speed.

YOLOv12 achieves both lower latency than previous YOLO models and higher accuracy when validated on the Microsoft COCO dataset. See the object detection model leaderboard for more details.

You can use the model for:

• Object detection
• Segmentation
• Classification
• Keypoint detection
• Oriented bounding box detection

You can see the model source code in the official model GitHub repository.