The generation of the textual description of the differences in images is a relatively new concept that requires the fusion of both computer vision and natural language techniques. In this paper, we present a novel Fully Convolutional CaptionNet (FCC) that employs an encoder-decoder framework to perform visual feature extractions, compute the feature distances, and generate new sentences describing the measured distances. After extracting the features of the images, a contrastive function is used to compute their weighted L1 distance which is learned and selectively attended to determine salient sections of the feature at every time step. The attended feature region is adequately matched to corresponding words iteratively until a sentence is completed. We propose the application of upsampling network to enlarge the features' field of view, this provides a robust pixel-based discrepancy computation. Our extensive experiments indicate that the FCC model outperforms other learning models on the benchmark Spot-the-Diff datasets by generating succinct and meaningful textual differences in images. INDEX TERMSImage captioning, deep learning, Siamese network, recurrent neural network, convolutional neural network, attention, fully convolutional networks.

No abstract

In Distributed Hash Table (DHT)-based Mobile Ad Hoc Networks (MANETs), a logical structured network (i.e., follows a tree, ring, chord, 3D, etc., structure) is built over the ad hoc physical topology in a distributed manner. The logical structures guide routing processes and eliminate flooding at the control and the data plans, thus making the system scalable. However, limited radio range, mobility, and lack of infrastructure introduce frequent and unpredictable changes to network topology, i.e., connectivity/dis-connectivity, node/link failure, network partition, and frequent merging. Moreover, every single change in the physical topology has an associated impact on the logical structured network and results in unevenly distributed and disrupted logical structures. This completely halts communication in the logical network, even physically connected nodes would not remain reachable due to disrupted logical structure, and unavailability of index information maintained at anchor nodes (ANs) in DHT networks. Therefore, distributed solutions are needed to tolerate faults in the logical network and provide end-to-end connectivity in such an adversarial environment. This paper defines the scope of the problem in the context of DHT networks and contributes a Fault-Tolerant DHT-based routing protocol (FTDN). FTDN, using a cross-layer design approach, investigates network dynamics in the physical network and adaptively makes arrangements to tolerate faults in the logically structured DHT network. In particular, FTDN ensures network availability (i.e., maintains connected and evenly distributed logical structures and ensures access to index information) in the face of failures and significantly improves performance. Analysis and simulation results show the effectiveness of the proposed solutions.

No abstract

Due to widespread availability of WiFi networks in buildings, indoor location based systems become a reality. To provide indoor location based services (LBS), finding the current location of a human, a computer, a mobile device or equipment such as a small UAV (like Quadcopter) is of great interest. The most prominent method for this purpose is the received signal strength (RSS)-based location from WiFi Access Points (APs) inside a building. Considerable amount of research is carried out in estimating current location and providing different services based on different wireless technologies. On the other hand, little attention has been paid to study and analyze the behavior of the received signal strength itself. It is challenging because the intensity of signal can change very frequently due to environmental features like topology, temperature, interaction with objects etc. In this paper, we study the behavior of WiFi signals in an indoor environment for RSS based localization by analyzing signals from three different access points by using triangulation technique. Our method is based on fingerprinting method. The experimental results reveal that the behavior of signal changes very frequently. The results lead us to the conclusion that understanding signals behavior is important before estimating current location and providing different LBS.