A large amount of multimedia data is required for applications with real-time monitoring needs. Wireless Multimedia Sensor Networks (WMSNs) have become the essential infrastructure for such applications. However, limitations such as different node capabilities, variable ambient conditions, and limited energy resources, make it difficult to achieve an ideal coverage in the mentioned networks. An optimization method that utilizes improved deep learning to enhance both the coverage performance and the operational lifetime of Wireless Multimedia Sensor Networks (WMSNs) was proposed. Convolutional Neural Networks (CNN) were used in the proposed method to extract the spatial information. In addition, a Deep Reinforcement Learning (DRL) agent was used to dynamically adjust the position of the sensor and routing strategies in response to feedback from the environment. The proposed model learned the optimal rules that increased the network coverage while minimizing the transmission latency and energy consumption by constructing coverage improvement as a Markov Decision Process (MDP). The proposed system provided a coverage ratio of 93.6%, energy consumption of 0.33, network lifetime of 945 nodes, packet delivery ratio of 92.1%, and end-to-end latency of 95 ms; thus outperforming existing methods like PSO and GA.