• IEMEK Journal of Embedded Systems and Applications
• /
• v.12 no.2
• /
• pp.113-120
• /
• 2017

In this paper, we present a night-time vehicle detection method using CNN (Convolutional Neural Network) classification. The camera based night-time vehicle detection plays an important role on various advanced driver assistance systems (ADAS) such as automatic head-lamp control system. The method consists mainly of thresholding, labeling and classification steps. The classification step is implemented by existing CIFAR-10 model CNN. Through the simulations tested on real road video, we show that CNN classification is a good alternative for night-time vehicle detection.

• International journal of advanced smart convergence
• /
• v.5 no.3
• /
• pp.1-7
• /
• 2016

In existing Convolutional Neural Network (CNNs) for object recognition task, there are only few efforts known to reduce the noises from the images. Both convolution and pooling layers perform the features extraction without considering the noises of the input image, treating all pixels equally important. In computer vision field, there has been a study to weight a pixel importance. Seam carving resizes an image by sacrificing the least important pixels, leaving only the most important ones. We propose a new way to combine seam carving approach with current existing CNN model for object recognition task. We attempt to remove the noises or the "unimportant" pixels in the image before doing convolution and pooling, in order to get better feature representatives. Our model shows promising result with CIFAR-10 dataset.

• Journal of Broadcast Engineering
• /
• v.26 no.7
• /
• pp.862-867
• /
• 2021

Although Deep Neural Networks (DNNs) have shown remarkable performance in various artificial intelligence fields, it is well known that DNNs are vulnerable to adversarial attacks. Since adversarial attacks are implemented by adding perturbations onto benign examples, increasing the sparsity of DNNs minimizes the propagation of errors to high-level layers. In this paper, unlike the traditional pruning scheme removing low magnitude weights, we eliminate high magnitude weights that are usually considered high absolute values, named 'reverse pruning' to ensure robustness. By conducting both theoretical and experimental analyses, we observe that reverse pruning ensures the robustness of DNNs. Experimental results show that our reverse pruning outperforms previous work with 29.01% in Top-1 accuracy on perturbed CIFAR-10. However, reverse pruning does not guarantee benign samples. To relax this problem, we further conducted experiments by adding a regularization term for the high magnitude weights. With adding the regularization term, we also applied conventional pruning to ensure the robustness of DNNs.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.12
• /
• pp.1665-1670
• /
• 2018

A convolutional neural network (CNN), which is one of the deep learning models, has been very successful in a variety of computer vision tasks. Filters of a CNN are automatically generated, however, they can be further optimized since there exist the possibility of existing redundant and less important features. Therefore, the aim of this paper is a filter reduction to accelerate and compress CNN models. Evolutionary algorithms is adopted to remove the unnecessary filters in order to minimize the parameters of CNN networks while maintaining a good performance of classification. We demonstrate the proposed filter reduction methods performing experiments on CIFAR10 data based on the classification performance. The comparison for three approaches is analysed and the outlook for the potential next steps is suggested.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.9
• /
• pp.3243-3257
• /
• 2021

Deep neural networks provide excellent performance in pattern recognition, audio classification, and image recognition. It is important that they accurately recognize input data, particularly when they are used in autonomous vehicles or for medical services. In this study, we propose a data correction method for increasing the accuracy of an unknown classifier by modifying the input data without changing the classifier. This method modifies the input data slightly so that the unknown classifier will correctly recognize the input data. It is an ensemble method that has the characteristic of transferability to an unknown classifier by generating corrected data that are correctly recognized by several classifiers that are known in advance. We tested our method using MNIST and CIFAR-10 as experimental data. The experimental results exhibit that the accuracy of the unknown classifier is a 100% correct recognition rate owing to the data correction generated by the proposed method, which minimizes data distortion to maintain the data's recognizability by humans.

• The Journal of Korea Robotics Society
• /
• v.17 no.2
• /
• pp.133-141
• /
• 2022

This paper introduces model compression algorithms which make a deep neural network smaller and faster for embedded systems. The model compression algorithms can be largely categorized into pruning, quantization and knowledge distillation. In this study, gradual pruning, quantization aware training, and knowledge distillation which learns the activation boundary in the hidden layer of the teacher neural network are integrated. As a large deep neural network is compressed and accelerated by these algorithms, embedded computing boards can run the deep neural network much faster with less memory usage while preserving the reasonable accuracy. To evaluate the performance of the compressed neural networks, we evaluate the size, latency and accuracy of the deep neural network, DenseNet201, for image classification with CIFAR-10 dataset on the NVIDIA Jetson Xavier.

• International Journal of Internet, Broadcasting and Communication
• /
• v.15 no.3
• /
• pp.66-72
• /
• 2023

Little attention appears to have been paid to the relevance of learning a good representation function in solving long tail tasks. Therefore, we propose a new loss function to ensure a good representation is learnt while learning to classify. We call this loss function Triplet Class-Wise Difficulty-Based (TriCDB-CE) Loss. It is a combination of the Triplet Loss and Class-wise Difficulty-Based Cross-Entropy (CDB-CE) Loss. We prove its effectiveness empirically by performing experiments on three benchmark datasets. We find improvement in accuracy after comparing with some baseline methods. For instance, in the CIFAR-10-LT, 7 percentage points (pp) increase relative to the CDB-CE Loss was recorded. There is more room for improvement on Places-LT.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2020.11a
• /
• pp.850-853
• /
• 2020

gradient decent 를 기반으로 한 Differentiable architecture search(DARTS)는 한 번의 Architecture Search 로 모든 후보 연산 중 가장 가중치가 높은 연산 하나를 선택한다. 이 때 비슷한 종류의 연산이 가중치를 나누어 갖는 "표의 분산"이 나타나, 성능이 더 좋은 연산이 선택되지 못하는 상황이 발생한다. 본 연구에서는 이러한 상황을 막기위해 Architecture Parameter 가중치의 gradient 를 기반으로 연산들을 클러스터링 하여 그룹화 한다. 그 후 그룹별로 가중치를 합산하여 높은 가중치를 갖는 그룹만을 사용하여 한 번 더 Architecture Search 를 진행한다. 각각의 Architecture Search 는 DARTS 의 절반 epoch 만큼 이루어지며, 총 epoch 이 같으나 두번째의 Architecture Search 는 선별된 연산 그룹을 사용하므로 DARTS 에 비해 더 적은 Search Cost 가 요구된다. "표의 분산"문제를 해결하고, 2 번으로 나뉜 Architecture Search 에 따라 CIFAR 10 데이터 셋에 대해 2.46%의 에러와 0.16 GPU-days 의 탐색시간을 얻을 수 있다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.580-581
• /
• 2024

고령화 사회에 접어들면서 황반 변성과 당뇨 망막 병증 등 시야결손을 동반하는 안구 질환의 발병률은 증가하지만 이러한 질환의 조기 발견에 인공지능을 접목시킨 연구는 부족한 실정이다. 본 논문은 안구 질환 자가 검사용 인공 신경망을 학습시키기 위한 데이터 베이스 구축 방법을 제안한다. MNIST와 CIFAR-10을 합성하여 중첩 이미지 데이터셋인 G-Dataset을 생성하였고, 7개의 인공신경망에 학습시켜 최종적으로 90% 이상의 정확도를 얻음으로 그 유효성을 입증하였다. G-Dataset을 안구 질환 자가 검사용 딥러닝 모델에 학습시켜 모바일 어플에 적용하면 사용자가 주기적인 검사를 통해 안구 질환을 조기에 진단하고 치료할 수 있을 것으로 기대된다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2017.06a
• /
• pp.243-244
• /
• 2017

최근 이미지 처리 및 인식 문제를 해결하는데 많이 사용되고 있는 CNN(Convolution Neural Network)를 이용하여 작은 dataset에서 Overfitting을 감소시키며 학습 할 수 있는 방법인 Dropout과 이미지를 왜곡하여 data를 늘리는 방법을 사용하여 보다 효율적으로 학습할 수 있는 방법을 연구 하였다. Batch별 처리속도를 기준으로 두 네트워크의 구조를 다르게 구현하여 비슷한 처리 시간을 수행하게 되도록 실험환경을 만들고 진행 하였다. Tensorflow로 네트워크를 구성하였고. Dataset은 Cifar_10을 사용 한다. 실험결과에 의하면 dropout의 경우 더 빨리 정확도가 향상되지만 이미지 왜곡을 사용하는 경우 저 높은 정확도로 수렴하였다.