Object detection automatically identifies and locates specific objects within images or videos, and plays a critical role in various fields. Among the many approaches, the YOLO model has significantly improved the accuracy and speed of object detection, thereby garnering attention for real-time processing. However, it exhibits limitations in detecting small or densely packed objects. This study introduces YOLO-RACE (You Only Look Once with ReAssembly and Convolutional block attEntion), an improved model based on YOLOv8 by incorporating the CARAFE upsampling technique, ResBlock, and the Convolutional Block Attention Module. YOLOv8 was selected from among various YOLO models because of its ability to deliver high accuracy in complex object detection tasks while maintaining its lightweight architecture and overall efficiency. The proposed model was evaluated using the VisDrone 2019, Pascal VOC, and SKU-110 K datasets. The experimental results demonstrated a precision of 0.419, recall of 0.321, mAP50 of 0.316, mAP50@95 of 0.18, and F1-score of 0.364 on the VisDrone 2019 dataset. On the Pascal VOC dataset, the model achieved a precision of 0.764, recall of 0.699, mAP50 of 0.773, mAP50@95 of 0.559, and F1-score of 0.73. Furthermore, on the SKU-110 K dataset, the model attained a precision of 0.904, recall of 0.841, mAP50 of 0.902, mAP50@95 of 0.575, and F1-score of 0.871. Based on these comparative results, the proposed model achieved superior performance compared with YOLOv8 and its variants, as well as with other state-of-the-art models such as YOLOv9, YOLOv10, YOLOv11, GELAN, and EfficientDet. This code is available at https://github.com/AnBLab-BAE/YOLO-RACE.git .

Abstract Generative Adversarial Networks (GAN) show excellent performance in various problems of computer vision, computer graphics, and machine learning, but require large amounts of data and huge computational resources. There is also the issue of unstable training. If the generator and discriminator diverge during the training process, the GAN is subsequently difficult to converge. In order to tackle these problems, various transfer learning methods have been introduced; however, mode collapse, which is a form of overfitting, often arises. Moreover, there were limitations in learning the distribution of the training data. In this paper, we provide a comprehensive review of the latest transfer learning methods as a solution to the problem, propose the most effective method of fixing some layers of the generator and discriminator, and discuss future prospects. The model to be used for the experiment is StyleGAN, and the performance evaluation uses Fréchet Inception Distance (FID), coverage, and density. Results of the experiment revealed that the proposed method did not overfit. The model was able to learn the distribution of the training data relatively well compared to the previously proposed methods. Moreover, it outperformed existing methods at the Stanford Cars, Stanford Dogs, Oxford Flower, Caltech-256, CUB-200–2011, and Insect-30 datasets.

QR code, for its recognition robustness and data capacity, has been often used for Augmented Reality applications as well as for other commercial applications. However, it is difficult to enable tangible interactions through which users may change 3D models or animations. It is because QR codes are automatically generated by the rules, and are not easily modifiable. Our goal was to enable QR code based Augmented Reality interactions. By analysis and through experiments, we discovered that some parts of a QR code can be altered to change the text string that the QR code represents. In this paper, we introduced a prototype for QR code based Augmented Reality interactions, which allows for Rubik's cube style rolling interactions.

QR code, for its recognition robustness and data capacity, has been often used for Augmented Reality applications as well as for other commercial applications. However, it is difficult to enable tangible interactions through which users may change 3D models or animations. It is because QR codes are automatically generated by the rules, and are not easily modifiable. Our goal was to enable QR code based Augmented Reality interactions. By analysis and through experiments, we discovered that some parts of a QR code can be altered to change the text string that the QR code represents. In this demo, we introduced a prototype for QR code based Augmented Reality interactions, which allows for Rubik's cube style rolling interactions.

New this year to ISMAR 2013, we are proud to present the Works In Progress (WIP) Program. Augmented Reality is rapidly growing into many new areas, so the WIP is a platform to present the field's latest, emerging results to the larger community before the work has reached its final form. This year, the program includes bread and butter AR technologies such as remote collaboration interfaces, fiducial marker design, and perceptual studies, to even loftier applications like AR interactions aboard the International Space Station. Passive haptics, bare-handed gesture interfaces, and realistic rendering round out the offerings. So come to the WIP sessions to hear about active AR research and find the spark of inspiration!

With popularization of smart phones, needs for location based services (LBS), which is one of the most promising Augmented Reality applications, increased rapidly. However, accuracy of most commercially available Global Positioning Systems (GPS) is below levels for providing practically meaningful location based information. Especially when there are high building structures nearby, GPS location measurements are known to be erroneous and deviant. In this paper, we present a computer vision based method for improving user's position and orientation for outdoor Augmented Reality with initial values obtained from a GPS and a digital compass. Given a digital map, our goal was to determine corresponding buildings visible in the camera image and improve the user location and orientation. In average, our method improved (14.4m, 3.3m) in position and 2.8 degrees in orientation. Our method is suitable for mobile services in urban environments where tall buildings degrade GPS signals.

In current Computer Aided Design (CAD) systems, designers are commonly restricted to a traditional workstation environment with mouse and keyboard. This environment is indirect from the physical object they are designing, and as such they may lose the one to one correspondence between the virtual and physical magnification of the design. In order to reduce this, we propose a Spatial Augmented Reality (SAR) based CAD system which consists of a fixed camera-projector pair, a Light Emitting Diode (LED) pen with two buttons, a wireless communication module, and a physical drawing board.

The S&T Program Chairs are delighted to welcome you to ISMAR 2011, the 10th symposium on Mixed and Augmented Reality! This year's symposium continues a long tradition of ISMAR meetings, a series that itself followed a related series of IWAR, ISMR, and ISAR meetings. This year's Science and Technology (S&T) track comprises a mix of highly selective mixed and augmented reality research and related work. Specifically, in the S&T program this year you will find 26 papers, 27 posters, as well as a stimulating mixture of keynote talks, demonstrations, tutorials, workshops and the tracking competition. All of these elements of the program are the result of dedicated hard work by members of various conference committees and additional volunteers, and we would like to thank all of them for their generous efforts.

Outdoor Augmented Reality (AR) gained popularity recently due to its potential for location based mobile services. However, most commercially available Global Positioning Systems (GPS), except for the expensive high-end models, do not provide accurate location information that is enough to be used for displaying practically meaningful location based information. In this paper, we present a computer vision based method for improving user's two dimensional location and one-dimensional orientation, the initial values of which are obtained from a GPS and a digital compass. Our method utilizes corner positions of buildings in the map and the vertical edges of the buildings in the captured images. We applied anisotropic diffusion in order to filter noise and preserve edges, and dual vertical edge filters on gray and saturation images. Our method is suitable for mobile services in urban environments where tall buildings degrade GPS signals. In average, our method improved 15.0 meters in position and 2.2 degrees in orientation.

This paper outlines the first Asian network-based speech-to-speech translation system developed by the Asian Speech Translation Advanced Research (A-STAR) consortium. The system was designed to translate common spoken utterances of travel conversations from a certain source language into multiple target languages in order to facilitate multiparty travel conversations between people speaking different Asian languages. Each A-STAR member contributes one or more of the following spoken language technologies: automatic speech recognition, machine translation, and text-to-speech through Web servers. Currently, the system has successfully covered 9 languages-namely, 8 Asian languages (Hindi, Indonesian, Japanese, Korean, Malay, Thai, Vietnamese, Chinese) and additionally, the English language. The system's domain covers about 20,000 travel expressions, including proper nouns that are names of famous places or attractions in Asian countries. In this paper, we discuss the difficulties involved in connecting various different spoken language translation systems through Web servers. We also present speech-translation results on the first A-STAR demo experiments carried out in July 2009.

Rapid prototyping allows for evaluation of design product in a short period of time. Designers or CEOs may perceive the size and appearance of the product by touching the automatically constructed physical objects produced through rapid prototyping. In this paper, we introduce 'augmented reality based reconfigurable mock-up', which enables interactive changes of shapes of products as well as colors, textures, and user interfaces. The shapes of the physical objects are reconfigurable by assembling different parts to the main body of the mock-up. Augmented reality technologies were used to alter the appearance of the mock-up by rendering the 3D models, colors, and interfaces accordingly. Developed AR-based reconfigurable mock-up is expected to be used for realistic design evaluation and CEO presentations.

Computer aided design applications have become designers' inevitable tools for expressing and simulating innovative ideas and concepts. However, replacing traditional materials and mock-ups with 3D CAD systems, designers are faced with the intangibility problem, unable to physically interact with test products in early stages of design process. As a touchable and graspable interface based on 3D CAD data, we propose augmented foam, which applies augmented reality technologies to physical blue foams. Using augmented foam, a blue foam mock-up is overlaid with a 3D virtual object, which is rendered with the same CAD model used for mock-up production. We presented a method to correct occlusions of the virtual products by user's hand. Augmented foam was tested for a mug design and a cleaning robot design. Designers were able to inspect and evaluate the design alternatives interactively and efficiently.

In early stages of production, storyboards are used for visually describing the story and the script. In This work, an augmented reality based storyboard-authoring tool is introduced. Proposed tool is easy-to-use, and provides intuitive interface for scene composition and camera pose/motion control. Using AR Storyboard, non-experienced users may compose 3D scenes for a Storyboard using interfaces in his/her real environments.

Outdoor augmented reality (AR) applications rely on hybrid tracking (GPS, digital compass, visual) for registration. RSC has developed a real-time visual tracking system that uses visual cues of buildings in an urban environment for correcting the results of a conventional tracking system. This approach relies on knowledge of a CAD model of the building. It not only provides motion estimation, but also absolute orientation/position. It is based on the "visual servoing" approach, originally developed for robotics tasks. We have demonstrated this approach in real-time at a building on the NRL campus This poster shows the approach and results. The concept can be generalized to any scenario where a CAD model is available. This system is being prepared for integration into the NRL system BARS (Battlefield Augmented Reality System).

This paper presents a Korean large vocabulary continuous speech recognition system based on pseudomorpheme units. In Korean, an eojeol (word phrase) is a unit for spacing and a morpheme is the smallest unit with semantic meaning. If the eojeol is used as the dictionary and language modeling unit, the number of the unit becomes enormous. Instead we propose to use modified morpheme or pseudomorpheme as the basic recognition unit. We can recover the original eojeol by concatenating graphemes of pseudomorpheme components. We used a dictionary and language model with pseudomorpheme/part-ofspeech entries where each entry can have multiple pronunciations according to the morphology rule. With 32k-word vocabulary, the speaker-independent character, pseudomorpheme, and eojeol recognition accuracies on economy article database were 90.8%, 84.5%, and 81.3%, respectively.

Phonetic transcriptions are often manually encoded in a pronunciation lexicon. This process is time consuming and requires linguistic expertise. Moreover, it is very difficult to maintain consistency. To handle these problems, we present a model that produces Korean pronunciation variants based on morphophonological analysis. By analyzing phonological variations frequently found in spoken Korean, we have derived about 800 phonemic contexts that would trigger the applications of the corresponding phonemic and allophonic rules. In generating pronunciation variants, morphological analysis is preceded to handle variations of phonological words. According to the morphological category, a set of finite state automata tables reflecting phonemic context is looked up to generate pronunciation variants. Our experiments show that the proposed model produces mostly correct pronunciation variants of phonological words consisting of several morphemes.

We have implemented a Korean-to-English and Korean-to-Japanese spoken language translation system prototype. The system can translate speech in the travel planning domain with 5,000 word vocabulary. In our prototype, we concentrate on how to transfer the intention of a user to the partner in spite of the current limitations of spoken language processing technology. We measured the end-to-end performance of the prototype to test whether the output of the system is understandable using a subjective measure. We also used an objective measure to evaluate the system performance and found that it generates coherent results with the subjective test. The test result shows that the user can understand the output even in the case where the system cannot translate speech correctly. Thus it is important to provide even partially correct translation output to the user, in order not to neglect the possibility that the user can infer the intended message using the context and his/her intelligence.

The registration problem is one of the major issues in augmented reality (AR). Fiducial tracking is gaining interest as a solution to this problem in video see-through AR because of the availability of digitized real scenes. There are several AR systems using fiducial tracking, but most of them operate in small desktop workspaces. It is difficult to apply them directly to large scale applications. The wide range of work distance and non-uniform lighting conditions make fiducial detection very difficult. Adding new fiducials requires off-line processing for measuring positions of new fiducials. We propose a fast and robust fiducial detection procedure with carefully designed color fiducials and noise analysis of digitized images. We also present a dynamic workspace extension method with on-line position determination of unknown features. We present a framework for applying AR to large scale applications.

This paper presents a new neural net ap- proach to robot visuo-motor coordination. The ap- proach, refered to here as neurobotics , establishes neural net-based visual error servoing as a fundamen- tal mechanism of implementing visuo-motor coordi- nation. Visual error servoing is achieved by project- ing the robot task space on the 3D perception net (representing the robot internal 3D space) and gen- erating an incremental change of 3D space arm con- figuration in the 3D perception net based on poten- tial field-based reactive path planning. The 3D space arm configuration planned in the 3D perception net is then translated into the corresponding joint angles through the arm kinematic net. The arm kinematic net is formed based on hierarchically self-organizing competitive and cooperative network. Simulation re- sults are shown.