Ta上传的资源 (0)

Bottleneck Transformers for Visual Recognition,在ImageNet上高达84.7%的top-1精度,性能优于SENet、EfficientNet等,将其应用于目标检测、实例分割等下游任务,涨点显著!