The early detection of skin cancer substantially improves the five-year survival rate of patients. It is often difficult to distinguish early malignant tumors from skin images, even by expert dermatologists. Therefore, several classification methods of dermatoscopic images have been proposed, but they have been found to be inadequate or defective for skin cancer detection, and often require a large amount of calculations. This study proposes an improved capsule network called FixCaps for dermoscopic image classification. FixCaps can obtain a larger receptive field than CapsNets by applying an efficient high-performance large-kernel at the bottom convolution layer whose kernel size is as large as 31 × 31, in contrast to commonly used 9 × 9. The convolutional block attention module was used to reduce the losses of spatial information caused by convolution and pooling. The group convolution was used to avoid model underfitting in the capsule layer. The network can improve the detection accuracy and reduce a great amount of calculations, compared with several existing methods. The experimental results showed that FixCaps is better than IRv2-SA in classification prediction of dermatoscopic images, which is achieved an accuracy of 96.49% on the HAM10000 dataset.