Fig. 2
From: An improved lightweight algorithm for traffic sign detection
SCC2f’s structure. This figure illustrates the architecture of the proposed SCC2f module. The input feature first passes through a 3\(\times\)3 convolution (SCConv), followed by a split operation and a series of lightweight bottleneck blocks. After feature concatenation and refinement, a second SCConv is applied. The output is further enhanced through two sequential branches: a Spatial Refinement Unit (SRU) and a Channel Refinement Unit (CRU), producing the final spatial-channel refined feature representation.
