In practical computer vision applications, such as autonomous driving, the ability to effectively process high-dynamic-range (HDR) scenes is crucial for safe operation. In this paper, we focus on object detection within HDR images. To address this, we propose a simple yet effective framework that employs multiple tone mappings. First, we generate multiple images from an HDR image with varying tone mapping parameters. Then, those images are fed into a high-performance object detector pre-trained with low-dynamic-range (LDR) images. Multiple detection results are merged with non-maximum suppression (NMS). To assess the performance of our method, we have built a validation dataset comprising HDR images captured in outdoor scenes with significant contrast variations. The experimental results using both our dataset and an existing one demonstrate that our method outperforms existing approaches1.1The code and the dataset can be available at https://open-vision.sc.e.titech.ac.jp/research/hdrdet.

On manipulation of objects for a person, body schema modification plays an important role to grasp the spatial relation between he/she and the object. In order to investigate the relation between body schema modification and performance of the spatial manipulation in this study, participants performed a task of tracing Japanese characters using sticks having different lengths, and their cerebral blood flow were measured during the task. Since body schema is a higher-order function of the brain, statistical parametric mapping (SPM) was used to elucidate the function. More over an improvement of the task performance by interference to the body schema by applying transcutaneous electrical nerve stimulation (TENS) with different intensities was tried, where the task performance means accuracy of the manipulation of object. Applying SPM analysis, a significant difference at Brodmann area 7 (BA7) and BA40 was confirmed when length of stick was the longest, which induced strongest body schema modification in our experimental condition. There was a significant difference in BA21 and BA39 when accuracy of the manipulation of object was different according to the length of the stick. Applying TENS to enhance the body schema modification, a significant different activation in BA19 and BA37 was confirmed. In conclusion, the brain part which related to accuracy of the hand manipulation were clarified. Moreover, we suggested TENS enable to change accuracy of the hand manipulation.

Desktop haptic device has been developed in the field of rehabilitation and entertainment. However, the desktop type restrains human's movement. Therefore, it is difficult to receive force sense information, moving to wide range position and posture. In this study, we developed a 1-DOF wearable haptic device with pneumatic artificial muscles and a MR brake. These smart actuators have high power density and change its output force structurally. Therefore, this haptic device can render various force sense such as elasticity, friction and viscosity. In this abstract, we describe two experiments rendering elasticity and friction to evaluate the performance of the device.

Recently, a robot's activity has been spreading into the care setting and homes. For these robots to work alongside humans requires them to have the same flexibility as that of the human muscle and ensure safety even if they collide with humans. To satisfy these requirements, we examined pneumatic artificial muscles—which are not only provide high output but are also light, flexible, and safe—as robot actuators. However, the commonly used McKibben-type artificial muscles have several drawbacks. Therefore, we developed straight-fiber-type artificial muscles, which have a contraction ratio and force that surpasses those of McKibben-type artificial muscles. We also developed a 6-DOF manipulator that is actuated by straight-fiber-type artificial muscles. However, since the DOF of this manipulator is less than that of the human arm, it is difficult for the end effector to make detailed movements. To solve this problem, we developed a 7-DOF manipulator, introducing a new mechanism in the joints. In this paper, we introduce a new restraint condition to the inverse kinematic computation of this manipulator and propose the method of redundancy use. Finally, we conduct an experiment following the trajectory and verify the effectiveness of the proposed method.

In mobile communications, people can be contacted by anyone, anywhere, and anytime, but this might interrupt them. We propose a new mobile telecommunication style, called mobile `Tsunagari-Kan', where users exchange situational cues constantly and casually so they can imagine the situation of the other party. To make this possible, we must decrease users' mental workload caused by frequent communications. We considered that the mental workload could be decreased by exchanging situational cues casually to achieve true `always-on' communications. To link the `flip action' of a cell phone with a communicational behavior, we have developed a test system called the Flip Communicator System based on the Tsunagari-kan concept. This prototype system enables users to exchange situational cues on their cell phones. We carried out a communication field test between seven sets of Japanese mothers and their children. The results indicate that our method can strengthen the ties between a mother and a child.

In this article, a new telecommunication environment for tsunagari-kan communication that enables users to exchange situational cues continuously and interactively via a network is described. In face-to-face communication or when communicators share one space physically, people exchange many situational cues. These cues allow one to sense or "feel" the other person, and they evoke emotional responses such as smiles and feelings of happiness. This tsunagari-kan communication aims at evoking thoughts about others in users' minds from cues or signs and at engendering tsunagari-kan (a sense of closeness to others), even when the individuals are in separate locations. This new style of telecommunication offers people comfort as if they are together, and it eventually helps them keep and foster their personal relationships. As the first step in verifying this concept, a communication terminal for family members living apart was developed, and a field test was run. This terminal facilitates the exchange of data about people's unconscious presence/motion cues and conscious touch signs. With an analysis of the field test results, the validity of this concept was confirmed.

The authors propose the fuzzy auto-regressive (AR) model and its applications. The identification and the estimation of the model and the model parameters are optimized by linear programming. The performance of the proposed model has already been tested by random data. They first propose an improved fuzzy AR model. The point of difference of the previous method is the objective function of the optimization. This method is applied to forecasting data of the living expenditure of a worker's household in Japan, and price index fuzzy time series. To use the fuzzy AR model, one can describe the behavior of fuzzy time series which cannot be described by the stochastic model.

A new parallel-processing wire-routing algorithm is presented and implemented on a parallel processor. The two main features of the parallel algorithm are the control of the path quality and the finding of a quasi-minimum Steiner tree. Both Lee's maze algorithm and the proposed algorithm are implemented on an AAP-1 two-dimensional array processor, and the performance is compared to that of software programming on a general-purpose computer. It is shown experimentally that routing by the proposed algorithm implemented on the AAP-1 is 230 times faster than a software maze router run on a 1-MIPS computer for a three-pin/net circuit on a 256 X 256 grid.

Presented is a new parallel processing wire routing algorithm that can control path quality in two point connections and find a quasi-minimum Steiner tree for multi-point connections. A parallel rip-up technique for finding a candidate path which blocks an unconnected net is also presented. These routing algorithms are implemented on a two-dimensional array processor, the AAP-1. It is experimentally shown that the average AAP-1 execution time per net using the routing algorithm is 100 msec in a 256X256 grid.