YOLOv6 has a series of models for various industrial scenarios, including N/T/S/M/L. Architectures vary considering the model size for better accuracy-speed trade-off. Bag-of-Freebies methods, such as self-distillation and additional training epochs, are introduced to further improve the performance. For industrial deployment, QAT with channel-wise distillation and graph optimization is used to pursue extreme performance. For more details, refer to https://arxiv.org/abs/2209.02976.
The following table lists the YOLOv6 detection model supported by the Vitis AI Library.