Jiwhan Kim

Salient Region Detection via High-Dimensional Color Transform

In this paper, we introduce a novel technique to automatically detect salient regions of an image via high-dimensional color transform. Our main idea is to represent a saliency map of an image as a linear combination of high-dimensional color space where salient regions and backgrounds can be distinctively separated. This is based on an observation that salient regions often have distinctive colors compared to the background in human perception, but human perception is often complicated and highly nonlinear. By mapping a low dimensional RGB color to a feature vector in a high-dimensional color space, we show that we can linearly separate the salient regions from the background by finding an optimal linear combination of color coefficients in the high-dimensional color space. Our high dimensional color space incorporates multiple color representations including RGB, CIELab, HSV and with gamma corrections to enrich its representative power. Our experimental results on three benchmark datasets show that our technique is effective, and it is computationally efficient in comparison to previous state-of-the-art techniques.

Single-Atom Catalyst of Platinum Supported on Titanium Nitride for Selective Electrochemical Reactions

As a catalyst, single-atom platinum may provide an ideal structure for platinum minimization. Herein, a single-atom catalyst of platinum supported on titanium nitride nanoparticles were successfully prepared with the aid of chlorine ligands. Unlike platinum nanoparticles, the single-atom active sites predominantly produced hydrogen peroxide in the electrochemical oxygen reduction with the highest mass activity reported so far. The electrocatalytic oxidation of small organic molecules, such as formic acid and methanol, also exhibited unique selectivity on the single-atom platinum catalyst. A lack of platinum ensemble sites changed the reaction pathway for the oxygen-reduction reaction toward a two-electron pathway and formic acid oxidation toward direct dehydrogenation, and also induced no activity for the methanol oxidation. This work demonstrates that single-atom platinum can be an efficient electrocatalyst with high mass activity and unique selectivity.

Deep Pyramidal Residual Networks

Deep convolutional neural networks (DCNNs) have shown remarkable performance in image classification tasks in recent years. Generally, deep neural network architectures are stacks consisting of a large number of convolutional layers, and they perform downsampling along the spatial dimension via pooling to reduce memory usage. Concurrently, the feature map dimension (i.e., the number of channels) is sharply increased at downsampling locations, which is essential to ensure effective performance because it increases the diversity of high-level attributes. This also applies to residual networks and is very closely related to their performance. In this research, instead of sharply increasing the feature map dimension at units that perform downsampling, we gradually increase the feature map dimension at all units to involve as many locations as possible. This design, which is discussed in depth together with our new insights, has proven to be an effective means of improving generalization ability. Furthermore, we propose a novel residual unit capable of further improving the classification accuracy with our new network architecture. Experiments on benchmark CIFAR-10, CIFAR-100, and ImageNet datasets have shown that our network architecture has superior generalization ability compared to the original residual networks. Code is available at https://github.com/jhkim89/PyramidNet.

Salient Region Detection via High-Dimensional Color Transform and Local Spatial Support

In this paper, we introduce a novel approach to automatically detect salient regions in an image. Our approach consists of global and local features, which complement each other to compute a saliency map. The first key idea of our work is to create a saliency map of an image by using a linear combination of colors in a high-dimensional color space. This is based on an observation that salient regions often have distinctive colors compared with backgrounds in human perception, however, human perception is complicated and highly nonlinear. By mapping the low-dimensional red, green, and blue color to a feature vector in a high-dimensional color space, we show that we can composite an accurate saliency map by finding the optimal linear combination of color coefficients in the high-dimensional color space. To further improve the performance of our saliency estimation, our second key idea is to utilize relative location and color contrast between superpixels as features and to resolve the saliency estimation from a trimap via a learning-based algorithm. The additional local features and learning-based algorithm complement the global estimation from the high-dimensional color transform-based algorithm. The experimental results on three benchmark datasets show that our approach is effective in comparison with the previous state-of-the-art saliency estimation methods.

Single-Atom Catalysts of Precious Metals for Electrochemical Reactions

Single-atom catalysts (SACs), in which metal atoms are dispersed on the support without forming nanoparticles, have been used for various heterogeneous reactions and most recently for electrochemical reactions. In this Minireview, recent examples of single-atom electrocatalysts used for the oxygen reduction reaction (ORR), hydrogen oxidation reaction (HOR), hydrogen evolution reaction (HER), formic acid oxidation reaction (FAOR), and methanol oxidation reaction (MOR) are introduced. Many density functional theory (DFT) simulations have predicted that SACs may be effective for CO2 reduction to methane or methanol production while suppressing H2 evolution, and those cases are introduced here as well. Single atoms, mainly Pt single atoms, have been deposited on TiN or TiC nanoparticles, defective graphene nanosheets, N-doped covalent triazine frameworks, graphitic carbon nitride, S-doped zeolite-templated carbon, and Sb-doped SnO2 surfaces. Scanning transmission electron microscopy, extended X-ray absorption fine structure measurement, and in situ infrared spectroscopy have been used to detect the single-atom structure and confirm the absence of nanoparticles. SACs have shown high mass activity, minimizing the use of precious metal, and unique selectivity distinct from nanoparticle catalysts owing to the absence of ensemble sites. Additional features that SACs should possess for effective electrochemical applications were also suggested.

One-pot synthesis of Pd@PdPt core–shell nanocubes on carbon supports

Cubic Pd@PdPt core–shell nanoparticles were synthesized on carbon supports using a one-pot process. The as-prepared cubic Pd@PdPt/C exhibited high electrocatalytic activity for oxygen reduction reactions. The enhancement in the mass activity increased after accelerated durability tests due to structural rearrangement.

Efficient and fast multi-view face detection based on feature transformation

The training time of Adaboost to obtain the strong classifier is usually time-consuming. Moreover, to deal with rotated faces, it is natural to need much more processing time for both training and execution stages. In this paper, we propose new efficient and fast multi-view face detection method based on Adaboost. From the robustness property of Harr-like feature, we first construct the strong classifier more effective to detect rotated face, and then we also propose new method that can reduce the training time. We call the method feature transformation method, which rotates and reflects entire weak classifiers of the strong classifier to construct new strong classifiers. Using our proposed feature transformation method, elapsed training time decrease significantly. We also test our face detectors on real-time HD images, and the results show the effectiveness of our proposed method.

Shaped platinum nanoparticles directly synthesized inside mesoporous silica supports

It is difficult to deposit shape-controlled nanoparticles into a mesoporous framework while preserving the shape. For shaped platinum nanoparticles, which are typically 5-10 nm in size, capillary inclusion by sonication or the formation of a mesoporous framework around the shaped platinum nanoparticles has been attempted, but the nanoparticles aggregated or their shapes were degraded easily. In this work, we directly nucleated platinum on the surface inside a mesoporous silica support and controlled the overgrowth step, producing cubic shaped nanoparticles. Mercaptopropyltrimethoxysilane was used as an anchoring agent causing nucleation at the silica surface, and it also helped to shape the nanoparticles. Platinum nanocubes, which were synthesized with polymeric capping agents separately, were deposited inside the mesoporous silica by sonication, but most of the nanoparticles were clogged at the entrance to the pores, and the surface of the platinum had very few sites that were catalytically active, as evidenced by the small H2 uptake. Unshaped platinum nanoparticles, which were prepared by conventional wet impregnation, showed a similar amount of H2 uptake as the in situ shaped platinum cubes, but the selectivity for pyrrole hydrogenation was poorer towards the production of pyrrolidine. The mesoporosity and the residual thiol groups on the surface of the in situ shaped Pt nanocubes might cause a high selectivity for pyrrolidine.

Automatic Image Segmentation Using Saliency Detection and Superpixel Graph Cuts

Image segmentation, which divides an image into foreground and background, is an important task for several applications in vision area such as object detection and classification. In this paper, we introduce a novel algorithm for automatic image segmentation technique which does not require further learning processes to perform segmentation. To achieve this automatic image segmentation, we incorporate saliency map for an image as an initial cue for image segmentation. An enhanced saliency detection method for generating saliency map is proposed. With over-segmented superpixels for an image and the generated saliency map, we perform image segmentation using graph cuts. To adapt graph cut segmentation to superpixel graph and saliency map, we suggest edge costs for superpixel graph based on Gaussian mixture models (GMM). As a result, superpixel graph enhances computational efficiency for our image segmentation technique and saliency map provides helpful cue for foreground region. We evaluate the performance of our algorithm on MSRA database demonstrate experimental results.

Facial age estimation via extended curvature Gabor filter

Facial age estimation is a process of identifying the age of a single face in an image or a video. Since age information can be used in many environments such as security, surveillance, and entertainment, age estimation has recently received much attention from researchers. In this paper, we propose an automatic age estimation method via extended curvature Gabor (ECG) features and a learning-based technique. Instead of conventional Gabor Filters, we use ECG filters to extract curvature information from a face image, which is useful for estimating age. We use a feature selection method to reduce the computational complexity and prove the effectiveness of ECG features at the same time. We use a regression algorithm to estimate the age of the test face image. As a result, our work achieves a competitive performance compared with other recent works in terms of age estimation.