The MobileNet class of Convolutional Neural networks (CNNs) is specifically designed for efficient execution on mobile and embedded devices with limited computing resources. An important feature of MobileNet is that it achieves higher classification accuracy with fewer parameters if it is designed on Lightweight Architecture (LA). To reduce the number of parameters and improve the classification accuracy, seed blocks are proposed in the Inception-Aided MobileNet model. The model is based on the Inception Module, which is inserted between MobileNet blocks to learn specific features. The initial block increased the number of parameters, but the iteration of the MobileNet block was reduced to balance the model parameters. The model presented in the paper, with fewer parameters and an initial block, performed better than the MobileNet model in the classification task. Experiments show that the model achieves higher recognition accuracy than MobileNet, with fewer parameters and lower computational costs.