SegNeXt: Rethinking Convolutional Attention Design for Semantic Segmentation

SegNeXt：重新思考语义分割中的卷积注意力设计.

Object-Contextual Representations for Semantic Segmentation

OCRNet：语义分割中的目标上下文表示.

Boundary loss for highly unbalanced segmentation

高度不平衡分割任务中的边界损失.

计算模型的参数量(Params)和运算量(FLOPs)

Compute the parameters and FLOPs of the model using thop.

Objects as Points

CenterNet：把目标检测视为中心点检测.

Argmax与SoftArgmax

Implement SoftArgmax with Pytorch.