Change captioning is the task of identifying the change and describing it with a concise caption. Despite recent advancements, filtering out insignificant changes still remains as a challenge. Namely, images from different camera perspectives can cause issues; a mere change in viewpoint should be disregarded while still capturing the actual changes. In order to tackle this problem, we present a new Viewpoint-Agnostic change captioning network with Cycle Consistency (VACC) that requires only one image each for the before and after scene, without depending on any other information. We achieve this by devising a new difference encoder module which can encode viewpoint information and model the difference more effectively. In addition, we propose a cycle consistency module that can potentially improve the performance of any change captioning networks in general by matching the composite feature of the generated caption and before image with the after image feature. We evaluate the performance of our proposed model across three datasets for change captioning, including a novel dataset we introduce here that contains images with changes under extreme viewpoint shifts. Through our experiments, we show the excellence of our method with respect to the CIDEr, BLEU-4, METEOR and SPICE scores. Moreover, we demonstrate that attaching our proposed cycle consistency module yields a performance boost for existing change captioning networks, even with varying image encoding mechanisms.

We discuss the globally optimal energy-efficient design of a 2D relay network. Different from the existing routing protocols on energy saving, which finds the minimal energy route for a given data rate, the proposed algorithm jointly optimizes routing and data rate to maximize energy efficiency (EE) defined as the achievable data rate per power consumption. We propose a low-complexity algorithm to circumvent the huge complexity of the exhaustive search for the network EE maximization and prove its global optimality. Moreover, the proposed algorithm is implemented in a distributed fashion because each relay needs to send its routing information only to the relays in its adjacent tiers, which significantly reduces the signaling overhead of the centralized implementation. Our analysis on the worst-case complexity in a fading channel shows that the complexity of the proposed algorithm increases linearly while that of the exhaustive search increases exponentially as the tier index increases. Simulation results confirm that the proposed algorithm outperforms the existing routing protocols on energy saving and achieves the globally optimal network EE at a significantly lower complexity than the exhaustive search.

We study energy-efficient power allocation for the sensing-based spectrum sharing (SBSS) of the cognitive users operating in an uplink multiuser MIMO mode. Different from the existing energy-efficient designs that consider system-wise energy efficiency, we consider user-wise energy efficiencies and optimize them in a Pareto sense. However, due to the non-convexity of the formulated problem, finding the optimal solution is a challenging problem. To resolve the non-convexity of the problem, we propose to reduce the feasible set by including additional power constraints. Also, we show that the feasible set reduction technique preserves the optimality of the original problem. Simulation results show that the proposed scheme significantly enhances the energy efficiency of the cognitive users compared with the existing spectral-efficient designs.

We study the energy efficiency issue in 5G communications scenarios, where cognitive femtocells coexist with picocells operating at the same frequency bands. Optimal energy-efficient power allocation based on the sensing-based spectrum sharing (SBSS) is proposed for the uplink cognitive femto users operating in a multiuser MIMO mode. Both hard-decision and soft-decision schemes are considered for the SBSS. Different from the existing energy-efficient designs in multiuser scenarios, which consider system-wise energy efficiency, we consider user-wise energy efficiency and optimize them in a Pareto sense. To resolve the nonconvexity of the formulated optimization problem, we include an additional power constraint to convexify the problem without losing global optimality. Simulation results show that the proposed schemes significantly enhance the energy efficiency of the cognitive femto users compared with the existing spectral-efficient designs.

In this paper, an optimal beamforming and power allocation for the sensing-based spectrum sharing (SBSS) that employs a soft decision on the state of the primary user is proposed. Different from the existing soft-decision-based SBSS that employs a single antenna and adjusts only transmit power, in the proposed scheme, the secondary transmitter employs multiple transmit antennas and controls the beamforming direction and the transmit power. We design the optimal beamforming and power allocation under the constraints of total transmit power and interference power, where both average and instantaneous power constraints are considered. Simulation results demonstrate that the proposed technique can significantly improve the average throughput of the secondary user compared with the existing methods. The proposed beamforming for the soft-decision-based SBSS is a promising scheme, particularly for cognitive femto base stations (BSs) to boost up the spectral efficiency while managing the interference to the macro BSs.

This paper proposes the eye detection using generalized binary pattern (GBP). The GBP can generate all possible patterns of ordered comparisons within a 3 × 3 neighborhood. Since existing local structure patterns consider all neighboring pixels around a given pixel, the number of possible patterns is fixed and limited to 2n. However, since the GBP takes the ordered comparisons of some partial neighboring pixels around a given pixel, a total of 502 different types can be generated in 3 × 3 block. So, our proposed GBP generates 19,162 binary patterns at the given pixel. Among the possible binary patterns, we take an effective set of pattern and position by the AdaBoost feature selection algorithm. Experimental results shows that the GBP provides higher eye detection accuracy on the BioID and FERET databases than other existing local structure patterns such as LBP and MCT.

Achieving computer vision on microscale devices is a challenge. On these platforms, the power and mass constraints are severe enough for even the most common computations (matrix manipulations, convolution, etc.) to be difficult. This paper proposes and analyzes a class of miniature vision sensors that can help overcome these constraints. These sensors reduce power requirements through template-based optical convolution, and they enable a wide field-of-view within a small form through a refractive optical design. We describe the tradeoffs between the field-of-view, volume, and mass of these sensors and we provide analytic tools to navigate the design space. We demonstrate milliscale prototypes for computer vision tasks such as locating edges, tracking targets, and detecting faces. Finally, we utilize photolithographic fabrication tools to further miniaturize the optical designs and demonstrate fiducial detection onboard a small autonomous air vehicle.

Diversity and transmission rate are the main performance metrics for MIMO techniques. We propose a low-complexity MIMO scheme that achieves both full diversity and full rate (FDFR) over flat fading channels when there is a sufficiently many number of transmit and receive antennas. The proposed scheme uses energy spreading transforms (EST's) to spread a symbol energy to multiple transmit antennas and makes iterative decision-feedback detection possible at the receiver. Simulation results show that the proposed scheme performs close to the (interference-free) genie-aided system when there are sufficiently large number of transmit and receive antennas.