• Journal of Korea Multimedia Society
• /
• v.23 no.8
• /
• pp.891-902
• /
• 2020

The Re-Identification(Re-ID) is one of the most popular researches in the field of computer vision due to a variety of applications. To achieve a high-level re-identification performance, recently other methods have developed the deep learning based networks that are specialized for only person or vehicle. However, most of the current methods are difficult to be used in real-world applications that require re-identification of both person and vehicle at the same time. To overcome this limitation, this paper proposes a deep neural network learning method that combines triplet and softmax loss to improve performance and re-identify people and vehicles simultaneously. It's possible to learn the detailed difference between the identities(IDs) by combining the softmax loss with the triplet loss. In addition, weights are devised to avoid bias in one-side loss when combining. We used Market-1501 and DukeMTMC-reID datasets, which are frequently used to evaluate person re-identification experiments. Moreover, the vehicle re-identification experiment was evaluated by using VeRi-776 and VehicleID datasets. Since the proposed method does not designed for a neural network specialized for a specific object, it can re-identify simultaneously both person and vehicle. To demonstrate this, an experiment was performed by using a person and vehicle re-identification dataset together.

• Journal of Information Processing Systems
• /
• v.15 no.4
• /
• pp.785-796
• /
• 2019

Due to the view point, illumination, personal gait and other background situation, person re-identification across cameras has been a challenging task in video surveillance area. In order to address the problem, a novel method called Joint Bayesian across different cameras for person re-identification (JBR) is proposed. Motivated by the superior measurement ability of Joint Bayesian, a set of Joint Bayesian matrices is obtained by learning with different camera pairs. With the global Joint Bayesian matrix, the proposed method combines the characteristics of multi-camera shooting and person re-identification. Then this method can improve the calculation precision of the similarity between two individuals by learning the transition between two cameras. For investigating the proposed method, it is implemented on two compare large-scale re-ID datasets, the Market-1501 and DukeMTMC-reID. The RANK-1 accuracy significantly increases about 3% and 4%, and the maximum a posterior (MAP) improves about 1% and 4%, respectively.

• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.89-106
• /
• 2022

Recently, the amount of video data collected from smartphones, CCTVs, black boxes, and high-definition cameras has increased rapidly. According to the increasing video data, the requirements for analysis and utilization are increasing. Due to the lack of skilled manpower to analyze videos in many industries, machine learning and artificial intelligence are actively used to assist manpower. In this situation, the demand for various computer vision technologies such as object detection and tracking, action detection, emotion detection, and Re-ID also increased rapidly. However, the object detection and tracking technology has many difficulties that degrade performance, such as re-appearance after the object's departure from the video recording location, and occlusion. Accordingly, action and emotion detection models based on object detection and tracking models also have difficulties in extracting data for each object. In addition, deep learning architectures consist of various models suffer from performance degradation due to bottlenects and lack of optimization. In this study, we propose an video analysis system consists of YOLOv5 based DeepSORT object tracking model, SlowFast based action recognition model, Torchreid based Re-ID model, and AWS Rekognition which is emotion recognition service. Proposed model uses single-linkage hierarchical clustering based Re-ID and some processing method which maximize hardware throughput. It has higher accuracy than the performance of the re-identification model using simple metrics, near real-time processing performance, and prevents tracking failure due to object departure and re-emergence, occlusion, etc. By continuously linking the action and facial emotion detection results of each object to the same object, it is possible to efficiently analyze videos. The re-identification model extracts a feature vector from the bounding box of object image detected by the object tracking model for each frame, and applies the single-linkage hierarchical clustering from the past frame using the extracted feature vectors to identify the same object that failed to track. Through the above process, it is possible to re-track the same object that has failed to tracking in the case of re-appearance or occlusion after leaving the video location. As a result, action and facial emotion detection results of the newly recognized object due to the tracking fails can be linked to those of the object that appeared in the past. On the other hand, as a way to improve processing performance, we introduce Bounding Box Queue by Object and Feature Queue method that can reduce RAM memory requirements while maximizing GPU memory throughput. Also we introduce the IoF(Intersection over Face) algorithm that allows facial emotion recognized through AWS Rekognition to be linked with object tracking information. The academic significance of this study is that the two-stage re-identification model can have real-time performance even in a high-cost environment that performs action and facial emotion detection according to processing techniques without reducing the accuracy by using simple metrics to achieve real-time performance. The practical implication of this study is that in various industrial fields that require action and facial emotion detection but have many difficulties due to the fails in object tracking can analyze videos effectively through proposed model. Proposed model which has high accuracy of retrace and processing performance can be used in various fields such as intelligent monitoring, observation services and behavioral or psychological analysis services where the integration of tracking information and extracted metadata creates greate industrial and business value. In the future, in order to measure the object tracking performance more precisely, there is a need to conduct an experiment using the MOT Challenge dataset, which is data used by many international conferences. We will investigate the problem that the IoF algorithm cannot solve to develop an additional complementary algorithm. In addition, we plan to conduct additional research to apply this model to various fields' dataset related to intelligent video analysis.

• International conference on construction engineering and project management
• /
• 2024.07a
• /
• pp.367-374
• /
• 2024

The detection and tracking of construction workers in building sites generate valuable data on unsafe behavior, work productivity, and construction progress. Many computer vision-based tracking approaches have been investigated and their capabilities for tracking construction workers have been tested. However, the dynamic nature of real-world construction environments, where workers wear similar outfits and move around in often cluttered and occluded regions, has severely limited the accuracy of these methods. Herein, to enhance the performance of vision-based tracking, a new framework is proposed which seamlessly integrates three computer vision components: detection, tracking, and re-identification (DTR). In DTR, a tracking algorithm continuously tracks identified workers using a detector and tracker in combination. Then, a re-identification model extracts visual features and utilizes them as appearance descriptors in subsequent frames during tracking. Empirical results demonstrate that the proposed method has excellent multi-object-tracking accuracy with better accuracy than an existing approach. The DTR framework can efficiently and accurately monitor workers, ensuring safer and more productive dynamic work environments.

• Annual Conference of KIPS
• /
• 2019.05a
• /
• pp.472-474
• /
• 2019

Because of the difference in network structure and loss function, Verification and identification models have their respective advantages and limitations for person reidentification (re-ID). In this work, we propose a multi-task network simultaneously computes the identification loss and verification loss for person reidentification. Given a pair of images as network input, the multi-task network simultaneously outputs the identities of the two images and whether the images belong to the same identity. In experiments, we analyze the major factors affect the accuracy of person reidentification. To address the occlusion problem and improve the generalization ability of reID models, we use the Random Erasing Augmentation (REA) method to preprocess the images. The method can be easily applied to different pre-trained networks, such as ResNet and VGG. The experimental results on the Market1501 datasets show significant and consistent improvements over the state-of-the-art methods.

• Korean Journal of Fisheries and Aquatic Sciences
• /
• v.49 no.2
• /
• pp.208-213
• /
• 2016

The re-identification of two Korean tonguefishes, Cynoglossus interruptus and Symphurus orientalis, was carried out using eight specimens collected from Korean waters in 2007 and 2013. C. interruptus is characterized by having a single row of scales between rows connected to the supraorbital line and the middle lateral line, 107–113 dorsal fin rays, 86–89 anal fin rays, and 53–55 vertebrae. S. orientalis is characterized by having a 1-2-2-2-2 ID pattern, 97–100 dorsal fin rays, 83–89 anal fin rays, and 52–55 vertebrae. Molecular analysis using mitochondrial DNA Cytochrome Oxidase subunit I sequences showed that specimens of the two species corresponded well to Japanese C. interruptus and Taiwanese S. orientalis, respectively. Therefore, although several reports have raised questions regarding the distribution of C. interruptus and S. orientalis in Korean waters, morphological and molecular data confirm that these two species are indeed distributed in these waters.

• Journal of Ecology and Environment
• /
• v.41 no.12
• /
• pp.345-350
• /
• 2017

Background: Ecological research often requires monitoring of a specific individual over an extended period of time. To enable non-invasive re-identification, consistent external marking is required. Treefrogs possess lateral lines for crypticity. While these patterns decrease predator detection, they also are individual specific patterns. In this study, we tested the use of lateral lines in captive and wild populations of Dryophytes japonicus as natural markers for individual identification. For the purpose of the study, the results of visual and software assisted identifications were compared. Results: In normalized laboratory conditions, a visual individual identification method resulted in a 0.00 rate of false-negative identification (RFNI) and a 0.0068 rate of false-positive identification (RFPI), whereas Wild-ID resulted in RFNI = 0.25 and RFNI = 0.00. In the wild, female and male data sets were tested. For both data sets, visual identification resulted in RFNI and RFPI of 0.00, whereas the RFNI was 1.0 and RFPI was 0.00 with Wild-ID. Wild-ID did not perform as well as visual identification methods and had low scores for matching photographs. The matching scores were significantly correlated with the continuity of the type of camera used in the field. Conclusions: We provide clear methodological guidelines for photographic identification of D. japonicus using their lateral lines. We also recommend the use of Wild-ID as a supplemental tool rather the principal identification method when analyzing large datasets.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.06a
• /
• pp.1329-1332
• /
• 2022

다중 객체 추적 기술은 스포츠, 문화 예술 공연, VR 등 여러 방송 콘텐츠에서 자주 사용되고 있다. 방송 영상 안에 등장하는 여러 객체들은 객체간 상호작용에 의해 가려짐, 사라짐 (Occlusion) 등의 현상이 빈번하게 발생하고, 이 경우 기존에 추적되어온 객체들의 ID 가 소실되거나 교환되는 문제가 발생한다. 본 논문에서 더 강인한 다중 객체 추적을 위해, 주 개체 뿐만 아니라 주 개체에 종속되는 하위 개체 또한 함께 추적하는 개체-쌍-추적 기법을 제안한다. 한 쌍으로 묶인 주 개체와 종속 개체의 추적 정보와 매칭 정보는 상호보완적으로 사용되어, 소실 및 교환된 ID 도 복원할 수 있는 가능성을 높일 수 있다. 본 논문에서는 재인지 성능 향상을 위한 개체 쌍 추적 기법을 기술하였고, 성능 평가를 통해 제안 방법이 재인지 성능 향상에 기여할 수 있음을 확인하였다.

• Cartoon and Animation Studies
• /
• s.42
• /
• pp.241-262
• /
• 2016

British TV Channel 4 is one of the famous TV Channel in the world. Its station ID has also played a leading role in the developments of Motion Graphics including station IDs. This station ID's main visual design concept is its name and an iconic logo '4' at the same time. The first channel 4 station ID was designed by using modular typography to construct the iconic '4'. Modular typography is a technique of creating letters with similar elements. Channel 4's station ID was constructed from coloured polygons. The polygons split and converge at the same point in 3D space. Modularity in Channel 4's station ID is evidenced by the similar units of polygons. After the first station ID, Channel 4 was re-branded. Eventhough the station IDs which followed did not use coloured and geometrical polygons, modularity is seen in most of the station IDs especially between 2004 - 2011. In these station IDs, the iconic '4' is formed from similar natural and environmental objects like rocks, buildings, lights etc. In this analysis paper, there is a visual narrative on the history of Channel 4, the concept of modular typography in the original station ID and the application of modular typography in other Channel 4's station IDs.

• International conference on construction engineering and project management
• /
• 2024.07a
• /
• pp.1311-1311
• /
• 2024

Recent advances in artificial intelligence technology have led to active research aimed at systematically managing the productivity and environmental impact of major management targets such as heavy equipment at construction sites. However, challenges arise due to phenomena like partial occlusions, resulting from the dynamic working environment of construction sites (e.g., equipment overlapping, obstruction by structures), which impose practical constraints on precisely monitoring heavy equipment. To address these challenges, this study aims to enhance automated multi-object tracking (MOT) in scenarios involving long-term occlusions across consecutive frames for heavy construction equipment. To achieve this, two methodologies are employed to address long-term occlusions at construction sites: (i) tracking-by-detection and (ii) video inpainting with generative adversarial networks (GANs). Firstly, this study proposes integrating FairMOT with a tracking-by-detection algorithm like ByteTrack or SMILEtrack, demonstrating the robustness of re-identification (Re-ID) in occlusion scenarios. This method maintains previously assigned IDs when heavy equipment is temporarily obscured and then reappears, analyzing location, appearance, or motion characteristics across consecutive frames. Secondly, adopting video inpainting with GAN algorithms such as ProPainter is proposed, demonstrating robustness in removing objects other than the target object (e.g., excavator) during the video preprocessing and filling removed areas using information from surrounding pixels or other frames. This approach addresses long-term occlusion issues by focusing on a single object rather than multiple objects. Through these proposed approaches, improvements in the efficiency and accuracy of detection, tracking, and activity recognition for multiple heavy equipment are expected, mitigating MOT challenges caused by occlusions in dynamic construction site environments. Consequently, these approaches are anticipated to play a significant role in systematically managing heavy equipment productivity, environmental impact, and worker safety through the development of advanced construction and management systems.