The transferring of huge multimedia data over the limited bandwidth environment has many challenges in real time. Wireless Multimedia Sensor Networks (WMSN) is a special type of wireless sensor networks which are used to overcome such bandwidth limitations in order to provide effective transferring of multimedia data. The malicious nodes in WMSN fail the links between the sensor nodes which degrades the efficiency of the entire network. Each node in WMSN may have its own signal transferring capability based on its energy level. If the energy level of the node degrades beyond the threshold level, that node becomes malicious node which is the main reason for the link failure between this node and its surrounding nodes. The data transfer is affected by the link failure nodes which degrades the performance of the entire system. Hence, the detection of link failure is important to improve the performance efficiency of the network system. This paper focuses the link failure detection system using deep learning approach. Hence, the detection of link failure is an important task to improve the performance of the network. This paper proposes an effective methodology for detecting the link failures of clusters in WMSN using deep learning architecture. The nodes in WMSN are grouped in to number of clusters and cluster head is determined using multi tier clustering approach, based on the energy levels and weighting metric approach. Then, the features are computed from each cluster and these features are classified using Visual Geometry Group (VGG) classification approach in order to detect the link failures of the clusters in WMSN. The performance of this developed methodology is analyzed with respect to Packet Delivery Ratio (PDR) and latency.