• ETRI Journal
• /
• v.42 no.2
• /
• pp.230-238
• /
• 2020

Human-object interaction (HOI) detection is a popular computer vision task that detects interactions between humans and objects. This task can be useful in many applications that require a deeper understanding of semantic scenes. Current HOI detection networks typically consist of a feature extractor followed by detection layers comprising small filters (eg, 1 × 1 or 3 × 3). Although small filters can capture local spatial features with a few parameters, they fail to capture larger context information relevant for recognizing interactions between humans and distant objects owing to their small receptive regions. Hence, we herein propose a three-stream HOI detection network that employs a context convolution module (CCM) in each stream branch. The CCM can capture larger contexts from input feature maps by adopting combinations of large separable convolution layers and residual-based convolution layers without increasing the number of parameters by using fewer large separable filters. We evaluate our HOI detection method using two benchmark datasets, V-COCO and HICO-DET, and demonstrate its state-of-the-art performance.

• Journal of Internet of Things and Convergence
• /
• v.6 no.3
• /
• pp.1-6
• /
• 2020

This paper aims to integrate the previous models about mean object transfer latency in one framework and analyze the result through the computational experience. The analytical object transfer latency model assumes the multiple packet losses and the Internet of Things(IoT) environment including multi-hop wireless network, where fast re-transmission is not possible due to small window. The model also considers the initial congestion window size and the multiple packet loss in one congestion window. Performance evaluation shows that the lower and upper bounds of the mean object transfer latency are almost the same when both transfer object size and packet loss rate are small. However, as packet loss rate increases, the size of the initial congestion window and the round-trip time affect the upper and lower bounds of the mean object transfer latency.

• 제어로봇시스템학회:학술대회논문집
• /
• 1996.10b
• /
• pp.1316-1319
• /
• 1996

This paper proposes the algorithm for the correction of a position and orientation of small object such as chip in the precise construction process. In the past, it is general to correct position and orientation of object using human sight and simple vision sensors. But recently, researches using image processing devices have been studied to improve the corrective precision of a position and orientation of object. In this piper, maximum-axis moment and p-theta algorithm are used to correct the position and orientation. Algorithm of maximum-axis moment is widely applied to hetero-object except being applied to a perfect rectangle. This is reason that moments of the X and Y-axis are equal. Therefore, being the shape of a perfect rectangle, the object is applied to other algorithm. In the light of time problem, real-time control is as important as correction of object. To solve it, we use the DSP(Digital Signal Processing) which is far more fast than PC.

• Proceedings of the IEEK Conference
• /
• 2003.11a
• /
• pp.233-236
• /
• 2003

In this paper, we propose a method to search the evidence such as a knife found in the crime scene based on GHT from an image database Such objects like knives are simitar in shape. The proposed method utilizes the small shape differences among objects as much as possible to distinguish an object from similar shaped objects. It consists of the GHT based candidate generation and top-down candidate verification. For the fast generation of the candidate 1ist, the GHT operation is performed un the down sampled edge list. The test results show that it can retrieve the correct object even with a pan of object in reasonable time.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.8
• /
• pp.1099-1109
• /
• 1995

By using a concept of interrelation quadruplet between line segments, a new method for recognition of object families is introduced. The interrelation quadruplet, which is invariant under similarity transform of a pair of line segments, is used as a feature information for polygonal shape recognition. Several useful propertes of the interrelation quadruplet are derived in relation to efficient recognition of object families. Compared with the previous methods, the proposed method requires only small space of storage and is shown to be computationally simple and efficient.

• The Journal of Korea Robotics Society
• /
• v.15 no.2
• /
• pp.184-189
• /
• 2020

This paper is a study on data augmentation for small dataset by using deep learning. In case of training a deep learning model for recognition and classification of non-mainstream objects, there is a limit to obtaining a large amount of training data. Therefore, this paper proposes a data augmentation method using perspective transform and image synthesis. In addition, it is necessary to save the object area for all training data to detect the object area. Thus, we devised a way to augment the data and save object regions at the same time. To verify the performance of the augmented data using the proposed method, an experiment was conducted to compare classification accuracy with the augmented data by the traditional method, and transfer learning was used in model learning. As experimental results, the model trained using the proposed method showed higher accuracy than the model trained using the traditional method.

• Journal of Korea Multimedia Society
• /
• v.22 no.2
• /
• pp.167-177
• /
• 2019

Even though so much progresses have been achieved in Multiple Object Tracking (MOT), most of reported MOT methods are not still satisfactory for commercial embedded products like Pan-Tilt-Zoom (PTZ) camera. In this paper, we propose a real-time multiple pedestrians tracking method for embedded environments. First, we design a new light weight convolutional neural network(CNN)-based pedestrian detector, which is constructed to detect even small size pedestrians, as well. For further saving of processing time, the designed detector is applied for every other frame, and Kalman filter is employed to predict pedestrians' positions in frames where the designed CNN-based detector is not applied. The pose orientation information is incorporated to enhance object association for tracking pedestrians without further computational cost. Through experiments on Nvidia's embedded computing board, Jetson TX2, it is verified that the designed pedestrian detector detects even small size pedestrians fast and well, compared to many state-of-the-art detectors, and that the proposed tracking method can track pedestrians in real-time and show accuracy performance comparably to performances of many state-of-the-art tracking methods, which do not target for operation in embedded systems.

• International journal of advanced smart convergence
• /
• v.11 no.2
• /
• pp.22-29
• /
• 2022

In this paper, for the purpose of designing an real-time unmanned monitoring system, the YOLOv5s (small) object detection model was applied on the NVIDIA TX2^TM AI (Artificial Intelligence) edge computing platform in order to design the fundamental function of an unmanned monitoring system that can detect objects in real time. YOLOv5s was applied to the our real-time unmanned monitoring system based on the performance evaluation of object detection algorithms (for example, R-CNN, SSD, RetinaNet, and YOLOv5). In addition, the performance of the four YOLOv5 models (small, medium, large, and xlarge) was compared and evaluated. Furthermore, based on these results, the YOLOv5s model suitable for the design purpose of this paper was ported to the NVIDIA TX2^TM AI edge computing system and it was confirmed that it operates normally. The real-time unmanned monitoring system designed as a result of the research can be applied to various application fields such as an security or monitoring system. Future research is to apply NMS (Non-Maximum Suppression) modification, model reconstruction, and parallel processing programming techniques using CUDA (Compute Unified Device Architecture) for the improvement of object detection speed and performance.

• International conference on construction engineering and project management
• /
• 2022.06a
• /
• pp.893-903
• /
• 2022

Recently, computer vision (CV)-based safety monitoring (i.e., object detection) system has been widely researched in the construction industry. Sufficient and high-quality data collection is required to detect objects accurately. Such data collection is significant for detecting small objects or images from different camera angles. Although several previous studies proposed novel data augmentation and synthetic data generation approaches, it is still not thoroughly addressed (i.e., limited accuracy) in the dynamic construction work environment. In this study, we proposed a game engine-driven synthetic data generation model to enhance the accuracy of the CV-based object detection model, mainly targeting small objects. In the virtual 3D environment, we generated synthetic data to complement training images by altering the virtual camera angles. The main contribution of this paper is to confirm whether synthetic data generated in the game engine can improve the accuracy of the CV-based object detection model.

• Proceedings of the KIEE Conference
• /
• 2003.11c
• /
• pp.400-403
• /
• 2003

This paper presents the signal processing of the panoramic thermal warning system that detects the small target such as aircraft and helicopter from afar. We develope the all round looking thermal imaging system which can scan all the way. This system acquires the panoramic images to reconstruct the IR images by revolving head of sensor typed line sensor at high speed. For detection, where the object of interest may be small, it is sometimes difficult to specify from object and background by conventional contrast enhancement methods. Therefore we use the adaptive plateau equalization algorithm each region to improve the contrast and make the hardware system which consists of the signal processing board for real-time display. We can verify the proposed method by the computer simulation and the hardware implementation.