Fig. 4 | Scientific Reports

Fig. 4

From: DSNet enables feature fusion and detail restoration for accurate object detection in foggy conditions

Fig. 4

The composition of the RCS module comprises: a the RepVGG structure utilized during the training phase and b the RepConv structure employed during model inference or deployment. In these structures, rectangles with black borders represent specific module operations performed on the tensors, while rectangles with gradient-filled shading indicate the properties of the tensors, with the width of the rectangle reflecting the number of channels in the tensor.

Back to article page