• Journal of the Korean Society of Industry Convergence
• /
• v.24 no.5
• /
• pp.547-553
• /
• 2021

Dynamic object recognition is an important task for autonomous vehicles. Since dynamic objects exhibit a higher collision risk than static objects, our own trajectories should be planned to match the future state of moving elements in the scene. Time information such as optical flow can be used to recognize movement. Existing optical flow calculations are based only on camera sensors and are prone to misunderstanding in low light conditions. In this regard, to improve recognition performance in low-light environments, we applied a normalization filter and a correction function for Gamma Value to the input images. The low light quality improvement algorithm can be applied to confirm the more accurate detection of Object's Bounding Box for the vehicle. It was confirmed that there is an important in object recognition through image prepocessing and deep learning using YOLO.

• International Journal of Computer Science & Network Security
• /
• v.21 no.8
• /
• pp.87-96
• /
• 2021

The demand for e-learning through video lectures is rapidly increasing due to its diverse advantages over the traditional learning methods. This led to massive volumes of web-based lecture videos. Indexing and retrieval of a lecture video or a lecture video topic has thus proved to be an exceptionally challenging problem. Many techniques listed by literature were either visual or audio based, but not both. Since the effects of both the visual and audio components are equally important for the content-based indexing and retrieval, the current work is focused on both these components. A framework for automatic topic-based indexing and search depending on the innate content of the lecture videos is presented. The text from the slides is extracted using the proposed Merged Bounding Box (MBB) text detector. The audio component text extraction is done using Google Speech Recognition (GSR) technology. This hybrid approach generates the indexing keywords from the merged transcripts of both the video and audio component extractors. The search within the indexed documents is optimized based on the Naïve Bayes (NB) Classification and K-Means Clustering models. This optimized search retrieves results by searching only the relevant document cluster in the predefined categories and not the whole lecture video corpus. The work is carried out on the dataset generated by assigning categories to the lecture video transcripts gathered from e-learning portals. The performance of search is assessed based on the accuracy and time taken. Further the improved accuracy of the proposed indexing technique is compared with the accepted chain indexing technique.

• Journal of Korea Multimedia Society
• /
• v.22 no.2
• /
• pp.136-145
• /
• 2019

The crowded environment of a domestic pig farm is highly vulnerable to the spread of infectious diseases such as foot-and-mouth disease, and studies have been conducted to automatically analyze behavior of pigs in a crowded pig farm through a video surveillance system using a camera. Although it is required to correctly separate occluding pigs for tracking each individual pigs, extracting the boundaries of the occluding pigs fast and accurately is a challenging issue due to the complicated occlusion patterns such as X shape and T shape. In this study, we propose a fast and accurate method to separate occluding pigs not only by exploiting the characteristics (i.e., one of the fast deep learning-based object detectors) of You Only Look Once, YOLO, but also by overcoming the limitation (i.e., the bounding box-based object detector) of YOLO with the test-time data augmentation of rotation. Experimental results with two-pigs occlusion patterns show that the proposed method can provide better accuracy and processing speed than one of the state-of-the-art widely used deep learning-based segmentation techniques such as Mask R-CNN (i.e., the performance improvement over Mask R-CNN was about 11 times, in terms of the accuracy/processing speed performance metrics).

• The Journal of Korea Robotics Society
• /
• v.14 no.2
• /
• pp.139-149
• /
• 2019

In this paper, we present auto-annotation tool and synthetic dataset using 3D CAD model for deep learning based object detection. To be used as training data for deep learning methods, class, segmentation, bounding-box, contour, and pose annotations of the object are needed. We propose an automated annotation tool and synthetic image generation. Our resulting synthetic dataset reflects occlusion between objects and applicable for both underwater and in-air environments. To verify our synthetic dataset, we use MASK R-CNN as a state-of-the-art method among object detection model using deep learning. For experiment, we make the experimental environment reflecting the actual underwater environment. We show that object detection model trained via our dataset show significantly accurate results and robustness for the underwater environment. Lastly, we verify that our synthetic dataset is suitable for deep learning model for the underwater environments.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.10
• /
• pp.3608-3626
• /
• 2021

The detection of bolts is an important task in high-speed train inspection systems, and it is frequently performed to ensure the safety of trains. The difficulty of the vision-based bolt inspection system lies in small sample defect detection, which makes the end-to-end network ineffective. In this paper, the problem is resolved in two stages, which includes the detection network and cascaded classification networks. For small bolt detection, all bolts including defective bolts and normal bolts are put together for conducting annotation training, a new loss function and a new boundingbox selection based on the smallest axis-aligned convex set are proposed. These allow YOLOv3 network to obtain the accurate position and bounding box of the various bolts. The average precision has been greatly improved on PASCAL VOC, MS COCO and actual data set. After that, the Siamese network is employed for estimating the status of the bolts. Using the convolutional Siamese network, we are able to get strong results on few-shot classification. Extensive experiments and comparisons on actual data set show that the system outperforms state-of-the-art algorithms in bolt inspection.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.12
• /
• pp.4816-4834
• /
• 2020

This paper proposes transfer learning and fine-tuning techniques for a deep learning model to detect three distinct brain tumors from Magnetic Resonance Imaging (MRI) scans. In this work, the recent YOLOv4 model trained using a collection of 3064 T1-weighted Contrast-Enhanced (CE)-MRI scans that were pre-processed and labeled for the task. This work trained with the partial 29-layer YOLOv4-Tiny and fine-tuned to work optimally and run efficiently in most platforms with reliable performance. With the help of transfer learning, the model had initial leverage to train faster with pre-trained weights from the COCO dataset, generating a robust set of features required for brain tumor detection. The results yielded the highest mean average precision of 93.14%, a 90.34% precision, 88.58% recall, and 89.45% F1-Score outperforming other previous versions of the YOLO detection models and other studies that used bounding box detections for the same task like Faster R-CNN. As concluded, the YOLOv4-Tiny can work efficiently to detect brain tumors automatically at a rapid phase with the help of proper fine-tuning and transfer learning. This work contributes mainly to assist medical experts in the diagnostic process of brain tumors.

• International Journal of Internet, Broadcasting and Communication
• /
• v.13 no.1
• /
• pp.54-60
• /
• 2021

AR (Augmented Reality) is a technology that provides virtual content to the real world and provides additional information to objects in real-time through 3D content. In the past, a high-performance device was required to experience AR, but it was possible to implement AR more easily by improving mobile performance and mounting various sensors such as ToF (Time-of-Flight). Also, the importance of mobile augmented reality is growing with the commercialization of high-speed wireless Internet such as 5G. Thus, this paper proposes a system that can provide AR services via GNN (Graph Neural Network) using cameras and sensors on mobile devices. ToF of mobile devices is used to capture depth maps. A 3D point cloud was created using RGB images to distinguish specific colors of objects. Point clouds created with RGB images and Depth Map perform downsampling for smooth communication between mobile and server. Point clouds sent to the server are used for 3D object detection. The detection process determines the class of objects and uses one point in the 3D bounding box as an anchor point. AR contents are provided through app and web through class and anchor of the detected object.

• Nuclear Engineering and Technology
• /
• v.54 no.9
• /
• pp.3242-3249
• /
• 2022

As Kori unit 1 is undergoing the decommissioning process, estimating the disposal amount of waste from the decommissioned nuclear reactor has become one of the challenging issues. Since the waste disposal amount estimation depends on the packing of the waste, it is highly desirable to optimize the waste packing plan. In this study, we developed an efficient scheme for packing waste component segments. The scheme consists of 1) preparing three-dimensional models of segments, 2) orienting each segment in such a way to minimize the bounding box volume, and 3) applying hybrid genetic algorithm to pack the segments in the disposal containers. When the packing solution converges in the algorithm, it comes up with the number of containers used and the placement of segments in each container. The scheme was applied to Kori-1 reactor pressure vessel. The required number of containers calculated by the developed scheme was 24 compared to 42 that was the estimation of the prior packing plan, resulting in disposal volume savings by more than 40%. The developed method is flexible for applications to various packing problems with waste segments from different cutting options and different sizes of containers.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.1
• /
• pp.245-265
• /
• 2022

In response to problems such as insufficient extraction information, low detection accuracy, and frequent misdetection in the field of Thangka image defects, this paper proposes a YOLOv5 prediction algorithm fused with the attention mechanism. Firstly, the Backbone network is used for feature extraction, and the attention mechanism is fused to represent different features, so that the network can fully extract the texture and semantic features of the defect area. The extracted features are then weighted and fused, so as to reduce the loss of information. Next, the weighted fused features are transferred to the Neck network, the semantic features and texture features of different layers are fused by FPN, and the defect target is located more accurately by PAN. In the detection network, the CIOU loss function is used to replace the GIOU loss function to locate the image defect area quickly and accurately, generate the bounding box, and predict the defect category. The results show that compared with the original network, YOLOv5-SE and YOLOv5-CBAM achieve an improvement of 8.95% and 12.87% in detection accuracy respectively. The improved networks can identify the location and category of defects more accurately, and greatly improve the accuracy of defect detection of Thangka images.

• Journal of Advanced Information Technology and Convergence
• /
• v.10 no.2
• /
• pp.167-175
• /
• 2020

In this paper, we propose a traffic accident detection on vehicle-mounted camera. In the proposed method, the minimum bounding box coordinates the central coordinates on the bird's eye view and motion vectors of each vehicle object, and ego-motions of the vehicle equipped with dash-cam are extracted from the dash-cam video. By using extracted 4 kinds features as the input of Bi-LSTM (bidirectional LSTM), the accident probability (score) is predicted. To investigate the effect of each input feature on the probability of an accident, we analyze the performance of the detection the case of using a single feature input and the case of using a combination of features as input, respectively. And in these two cases, different detection models are defined and used. Bi-LSTM is used as a cascade, especially when a combination of the features is used as input. The proposed method shows 76.1% precision and 75.6% recall, which is superior to our previous work.