• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.9
• /
• pp.349-358
• /
• 2021

Recently, a method of finding the attention area or localization area for an object of an image using CAM (Class Activation Map)[1] has been variously carried out as a study of WSOL (Weakly Supervised Object Localization). The attention area extraction from the object heat map using CAM has a disadvantage in that it cannot find the entire area of the object by focusing mainly on the part where the features are most concentrated in the object. To improve this, using CAM and Selective Search[6] together, we first expand the attention area in the heat map, and a Gaussian smoothing is applied to the extended area to generate retraining data. Finally we train the data to expand the attention area of the objects. The proposed method requires retraining only once, and the search time to find an localization area is greatly reduced since the selective search is not needed in this stage. Through the experiment, the attention area was expanded from the existing CAM heat maps, and in the calculation of IOU (Intersection of Union) with the ground truth for the bounding box of the expanded attention area, about 58% was improved compared to the existing CAM.

• Bulletin of the Korean Chemical Society
• /
• v.22 no.5
• /
• pp.488-492
• /
• 2001

As gas chromatography/infrared spectrometry (GC/IR) becomes routinely avaliable, methods must be developed to deal with the large amount of data produced. We demonstrate computer methods that quickly search through a large data file, locating thos e spectra that display a spectral feature of interest. Based on a modified library search routine, these selective data reduction methods retrieve all or nearly all of the compounds of interest, while rejecting the vast majority of unrelated compounds. To overcome the shifting problem of IR spectra, a search method of moving the average pattern was designed. In this moving pattern search, the average pattern of a particular functional group was not held stationary, but was allowed to be moved a little bit right and left.

• Journal of Electrical Engineering and Technology
• /
• v.10 no.3
• /
• pp.1227-1232
• /
• 2015

Several works help make sliding window object detection fast, nevertheless, computational demands remain prohibitive for numerous applications. This paper proposes a fast object detection method based on three strategies: cascade classifier, selective window search and fast feature extraction. Experimental results show that the proposed method outperforms the compared methods and achieves both high detection precision and low computation cost. Our approach runs at 17ms per frame on 640×480 images while attaining state-of-the-art accuracy.

• Journal of Broadcast Engineering
• /
• v.7 no.4
• /
• pp.317-326
• /
• 2002

The adaptive coding schemes in H.264 standardization provide a significant ceding efficiency and some additional features like error resilience and network friendliness. The variable block size motion compensation using multiple reference frames is one of the key H.264 coding elements to provide main performance gain, but also the main culprit that increases the overall computational complexity. For this reason, this paper proposes a selective motion estimation algorithm based on variable block size for fast motion estimation in H.264. After we find the SAD(Sum of Absolute Difference) at initial points using diamond search, we decide whether to perform additional motion search in each block. Simulation results show that the proposed method is five times faster than the conventional full search in case of search range $\pm$32.

• Korean Journal of Cognitive Science
• /
• v.16 no.1
• /
• pp.1-15
• /
• 2005

Since human visual system has limited capacity for visual information processing, it should select goal-relevant information for further processing. There have been several studies that emphasized the possible involvement of memory in spatial shift of selective attention (Chun & Jiang, 1998, 1999; Klein, 1988; Klein & MacInnes, 1999). However, other studies suggested the inferiority of human visual memory in change detection(Rensink, O'Regan, & Clark, 1997; Simons & Levin, 1997) and in visual search(Hotowitz & Wolfe, 1998). The present study examined the involvement of memory in visual search; whether memory for the previously searched items guides selective attentional shift or not. We investigated how search works by comparing visual search performances in three different conditions; full exposure condition, partial exposure condition, and partial-to-full exposure condition. Revisiting searched items was allowed only in full exposure condition and not in either partial or partial-to-full exposure condition. The results showed that the efficiencies of attentional shift were nearly identical for all conditions. This finding implies that even in full exposure condition the participants scarcely re-examined the previously searched items. The results suggest that instant memory can be formed and used in visual search process. These results disagree with the earlier studies claiming thar visual search has no memory. We discussed the problems of the previous research paradigms and suggested some alternative accounts.

• The KIPS Transactions:PartB
• /
• v.17B no.2
• /
• pp.149-156
• /
• 2010

Since the premature convergence phenomenon of genetic algorithms (GAs) degrades the performances of GAs significantly, solving this problem provides a lot of effects to the performances of GAs. In this paper, we propose a selective mutation method in order to improve the performances of GAs by alleviating this phenomenon. In the selective mutation, individuals are additionally mutated at the specific region according to their ranks. From this selective mutation, individuals with low ranks are changed a lot and those with high ranks are changed small in the phenotype. Finally, some good individuals search around them in detail and the other individuals have more chances to search new areas. This results in enhancing the performances of GAs through alleviating of the premature convergence phenomenon. We measured the performances of our method with four typical function optimization problems. It was found from experiments that our proposed method considerably improved the performances of GAs.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.4
• /
• pp.848-854
• /
• 2015

There are a lot of encryption methods to secure database proposed recently. Those encryption methods can protect the sensitive data of users effectively, but it deteriorates the search performance of database query. In this paper, we proposed the selective element encryption method in order to complement those drawbacks. In addition, we compared the performance of the proposed method with that of tuple level encryption method using the various queries. As a result, we found that the proposed method, which use the selective element encryption with bloom filter as a index, has better performance than the other encryption method.

• Journal of the Institute of Convergence Signal Processing
• /
• v.16 no.3
• /
• pp.83-89
• /
• 2015

Connected component labeling (CCL) is a mandatory step in image segmentation where objects are extracted and uniquely labeled. CCL is a computationally expensive operation and thus is often done in parallel processing framework to reduce execution time. Various parallel CCL methods have been proposed in the literature. Among them are NSZ label equivalence (NSZ-LE) method, modified 8 directional label selection (M8DLS) method, HYBRID1 method, and HYBRID2 method. Soh et al. showed that HYBRID2 outperforms the others and is the best so far. In this paper we propose a new hybrid parallel CCL algorithm termed as HYBRID3 that combines selective four directional label search (S4DLS) with label backtracking (LB). We show that the average percentage speedup of the proposed over M8DLS is around 60% more than that of HYBRID2 over M8DLS for various kinds of images.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2001.10a
• /
• pp.363-366
• /
• 2001

It is very difficult to maintain the performance of estimation models for the new breed of projects since the computing environment changes so rapidly in terms of programming languages, development tools, and methodologies. So, we propose to use the relevant cases for a neural network model, whose cost is the decreased number of cases. To balance the relevance and data availability, the qualitative input factors are used as criteria of data classification. With the data sets that have the same value for certain qualitative input factors, we can eliminate the factors from the model making reduced neural network models. So we need to seek the optimally reduced neural network model among them. To find the optimally case-selective neural network, we propose the search techniques and sensitivity analysis between data points and search space.

• Journal of Korea Multimedia Society
• /
• v.18 no.11
• /
• pp.1289-1301
• /
• 2015

Pedestrian detection (PD) is an essential task in various applications and sliding window-based methods utilizing HOG (Histogram of Oriented Gradients) or HOG-like descriptors have been shown to be very effective for accurate PD. However, due to exhaustive search across images, PD methods based on sliding window usually require heavy computational time. In this paper, we propose a real-time PD method for embedded visual surveillance with fixed backgrounds. The proposed PD method employs HOG descriptors as many PD methods does, but utilizes selective search so that it can save processing time significantly. The proposed selective search is guided by restricting searching to candidate regions extracted from Adaptive Gaussian Mixture Model (AGMM)-based background subtraction technique. Moreover, approximate computation of HOG descriptor and implementation in fixed-point arithmetic mode contributes to reduction of processing time further. Possible accuracy degradation due to approximate computation is compensated by applying an appropriate one among three offline trained SVM classifiers according to sizes of candidate regions. The experimental results show that the proposed PD method significantly improves processing speed without noticeable accuracy degradation compared to the original HOG-based PD and HOG with cascade SVM so that it is a suitable real-time PD implementation for embedded surveillance systems.