Table 2 FLOPs and param data for different models with the same input.

From: An improved lightweight tongue segmentation model with self-attention parallel network and progressive upsampling