The use of sensor networks is expanding due to the spread of the Internet of Things (IoT). As this expansion continues, the amount of data to be acquired will increase and the communication bandwidth may become compressed. In addition, the processing of the retrieved data is currently performed by the server, and deep learning is often used when processing the data. This process is heavy, and the load on the server increases as the amount of data increases. In order to reduce the load on the server, conventional research has proposed a method of distributed processing using edge computing and parallel processing using mobile devices. However, although the data processing speed is fast with these methods, there are problems of increased communication traffic and increased power consumption. Therefore, in this study, we propose a method of assigning intermediate layers for deep learning according to the processing capacity of each sensor for the purpose of reducing the traffic and server load in the wireless sensor network.