MODIFICATION OF ALEXNET ARCHITECTURE FOR DETECTION OF CAR PARKING AVAILABILITY IN VIDEO CCTV

The difficulty of finding a parking space in public places, especially during peak hours is a problem experienced by drivers. To assist the driver in finding parking space availability, a system is needed to monitor parking availability. One study to detect the availability of parking lots utilizing CCTV. However, research on the availability of parking spaces on CCTV data has several problems, detecting parking slots that are done manually to be inefficient when applied to different parking lots. Also, research to detect the availability of parking lots using the Convolution Neural Network (CNN) method with existing architecture has many parameters. Therefore, this study proposes a system to detect the availability of car parking lots using You Only Look Once (YOLO) V3 for marking the parking space and proposed a new architecture CNN called Lite AlexNet which has few parameters than other methods to speed up the process of detecting parking space availability. The best accuracy of the marking stage using YOLO V3 is 92.31% where the weather was cloudy. For the proposed Lite AlexNet get the best time training average which is 7 second compare to other existing methods and the average accuracy in every condition is 92.33% better than other methods.

Download Full-text

Sonar Image Target Detection and Recognition Based on Convolution Neural Network

Mobile Information Systems ◽

10.1155/2021/5589154 ◽

2021 ◽

Vol 2021 ◽

pp. 1-8

Author(s):

Wu Yanchen

Keyword(s):

Neural Network ◽

Recognition Task ◽

Network Models ◽

Convolution Neural Network ◽

Sonar Image ◽

Optical Images ◽

The Neural Network ◽

Average Accuracy ◽

Better Than ◽

Detection And Recognition

Recent advancements in deep learning offer an effective approach for the study in machine vision using optical images. In this paper, a convolution neural network is used to deal with the target task of sonar detection, and the performance of each neural network model in the sonar image detection and recognition task of underwater box and tire is compared. The simulation results show that the neural network method proposed in this paper is better than the traditional machine learning methods and SSD network models. The average accuracy of the proposed method for sonar image target recognition is 93%, and the detection time of a single image is only 0.3 seconds.

Download Full-text

Double attention recurrent convolution neural network for answer selection

Royal Society Open Science ◽

10.1098/rsos.191517 ◽

2020 ◽

Vol 7 (5) ◽

pp. 191517

Author(s):

Ganchao Bao ◽

Yuan Wei ◽

Xin Sun ◽

Hongli Zhang

Keyword(s):

Neural Network ◽

Question Answering ◽

Convolution Neural Network ◽

Deep Model ◽

Selection Tasks ◽

Matching Score ◽

Key Steps ◽

Chinese And English ◽

Design Inspiration ◽

Better Than

Answer selection is one of the key steps in many question answering (QA) applications. In this paper, a new deep model with two kinds of attention is proposed for answer selection: the double attention recurrent convolution neural network (DARCNN). Double attention means self-attention and cross-attention. The design inspiration of this model came from the transformer in the domain of machine translation. Self-attention can directly calculate dependencies between words regardless of the distance. However, self-attention ignores the distinction between its surrounding words and other words. Thus, we design a decay self-attention that prioritizes local words in a sentence. In addition, cross-attention is established to achieve interaction between question and candidate answer. With the outputs of self-attention and decay self-attention, we can get two kinds of interactive information via cross-attention. Finally, using the feature vectors of the question and answer, elementwise multiplication is used to combine with them and multilayer perceptron is used to predict the matching score. Experimental results on four QA datasets containing Chinese and English show that DARCNN performs better than other answer selection models, thereby demonstrating the effectiveness of self-attention, decay self-attention and cross-attention in answer selection tasks.

Download Full-text

On the Effect of Training Convolution Neural Network for Millimeter-Wave Radar-Based Hand Gesture Recognition

Sensors ◽

10.3390/s21010259 ◽

2021 ◽

Vol 21 (1) ◽

pp. 259

Author(s):

Kang Zhang ◽

Shengchang Lan ◽

Guiyuan Zhang

Keyword(s):

Neural Network ◽

Millimeter Wave ◽

Gesture Recognition ◽

Radar Data ◽

Hand Gesture Recognition ◽

Convolution Neural Network ◽

Hand Gesture ◽

Millimeter Wave Radar ◽

Average Accuracy ◽

Wave Radar

The purpose of this paper was to investigate the effect of a training state-of-the-art convolution neural network (CNN) for millimeter-wave radar-based hand gesture recognition (MR-HGR). Focusing on the small training dataset problem in MR-HGR, this paper first proposed to transfer the knowledge with the CNN models in computer vision to MR-HGR by fine-tuning the models with radar data samples. Meanwhile, for the different data modality in MR-HGR, a parameterized representation of temporal space-velocity (TSV) spectrogram was proposed as an integrated data modality of the time-evolving hand gesture features in the radar echo signals. The TSV spectrograms representing six common gestures in human–computer interaction (HCI) from nine volunteers were used as the data samples in the experiment. The evaluated models included ResNet with 50, 101, and 152 layers, DenseNet with 121, 161 and 169 layers, as well as light-weight MobileNet V2 and ShuffleNet V2, mostly proposed by many latest publications. In the experiment, not only self-testing (ST), but also more persuasive cross-testing (CT), were implemented to evaluate whether the fine-tuned models generalize to the radar data samples. The CT results show that the best fine-tuned models can reach to an average accuracy higher than 93% with a comparable ST average accuracy almost 100%. Moreover, in order to alleviate the problem caused by private gesture habits, an auxiliary test was performed by augmenting four shots of the gestures with the heaviest misclassifications into the training set. This enriching test is similar with the scenario that a tablet reacts to a new user. The results of two different volunteer in the enriching test shows that the average accuracy of the enriched gesture can be improved from 55.59% and 65.58% to 90.66% and 95.95% respectively. Compared with some baseline work in MR-HGR, the investigation by this paper can be beneficial in promoting MR-HGR in future industry applications and consumer electronic design.

Download Full-text

Pengembangan Deteksi Citra Mobil untuk Mengetahui Jumlah Tempat Parkir Menggunakan CUDA dan Modified YOLO

Jurnal Teknologi Informasi dan Ilmu Komputer ◽

10.25126/jtiik.2019641275 ◽

2019 ◽

Vol 6 (4) ◽

pp. 413 ◽

Cited By ~ 1

Author(s):

Sisco Jupiyandi ◽

Fadhil Rizqullah Saniputra ◽

Yoga Pratama ◽

Muhammad Robby Dharmawan ◽

Imam Cholissodin

Keyword(s):

Parallel Computing ◽

Computing Time ◽

Gpu Programming ◽

Parking Space ◽

Test Results ◽

Processing Data ◽

Average Accuracy ◽

Parking Lot ◽

Long Time ◽

Speed Up

Besarnya lahan pada parkir dan jumlah kendaraan roda empat dalam hal ini adalah mobil, dapat menjadi kendala bagi pengendara lain dalam mengetahui posisi parkir mana yang masih dapat digunakan. Sistem pengembangan perparkiran yang ada masih kurang maksimal dalam memanfaatkan lahan dan efisiensi waktunya. Berdasarkan banyaknya kendaraan mobil yang semakin bertambah, maka kebutuhan akan lahan parkir juga semakin dibutuhkan. Banyak sekali sistem yang belum dapat menangani berbagai permasalahan yang ada. Sistem ini dapat mengetahui jumlah slot pada lahan parkir dengan akurat sehingga memudahkan pengelola. Selain itu sistem ini juga dikembangkan agar waktu pencarian lahan parkir oleh pengguna parkir bisa sangat cepat. Sistem ini menggunakan penerapan pemrograman GPU yang dikombinasi dengan Modified Yolo (M-Yolo). GPU pada M-Yolo dibutuhkan untuk mengolah citra sekaligus mengolah data untuk mendeteksi citra mobil dan jumlah mobil secara paralel. Hasil uji coba menunjukkan bahwa dengan menggunakan GPU dibandingkan dengan CPU dapat mempercepat waktu komputasi rata-rata sebesar 0,179 detik dengan rata-rata akurasi sebesar 100%.AbstractThe width of parking lot and the number of cars in the parking lot can be an obstacle for motorists to know the parking area in which part is still empty. Parking systems that exist at this time are still not maximal in the utilization of parking lots and time efficiency. Based on the number of vehicles that are growing, then the need for parking space is also more needed. Many of the existing parking systems have not been able to handle the various problems. This system can know the number of slots on the parking lot, making it easier for operators to know the empty parking lot. In addition, this system will also be designed so that parking time search by parking users doesn’t take a long time. This system uses implementation of GPU programming mixed with Modified Yolo (M-Yolo). GPU on M-Yolo is needed to process images while processing data to detect car and the number of cars using parallel computing. The test results show that using the GPU compared to the CPU can speed up the average computing time by 0.179 seconds and it obtained an average accuracy of 100%.

Download Full-text

A Recognition Algorithm for Modulation Schemes by Convolution Neural Network and Spectrum Texture

Xibei Gongye Daxue Xuebao/Journal of Northwestern Polytechnical University ◽

10.1051/jnwpu/20193740816 ◽

2019 ◽

Vol 37 (4) ◽

pp. 816-823

Author(s):

Pengyuan Bai ◽

Hua Xu ◽

Li Sun

Keyword(s):

Neural Network ◽

Recognition Algorithm ◽

Convolution Neural Network ◽

Image Texture ◽

Communication Signals ◽

Softmax Classifier ◽

Spectrum Monitoring ◽

Traditional Algorithm ◽

Modulation Schemes ◽

Better Than

The recognition of modulation schemes for communication signals is an important part of communication surveillance and spectrum monitoring. An algorithm based on deep learning and spectrum texture is proposed to recognize modulation schemes. Based on imperceptible differences among various spectrums of modulation schemes, the algorithm uses Convolution Neural Network to capture the features of image texture and thus classify the features with a SOFTMAX classifier. The experiment shows the algorithm performs better than traditional algorithm based on feature parameters, while the features captured can better reveal the signal detail and reduces effort on feature parameter design.

Download Full-text

Pre-trained Deep Convolution Neural Network Model With Attention for Speech Emotion Recognition

Frontiers in Physiology ◽

10.3389/fphys.2021.643202 ◽

2021 ◽

Vol 12 ◽

Author(s):

Hua Zhang ◽

Ruoyun Gou ◽

Jili Shang ◽

Fangyao Shen ◽

Yifan Wu ◽

...

Keyword(s):

Neural Network ◽

Emotion Recognition ◽

Short Term Memory ◽

Convolution Neural Network ◽

Classification Model ◽

Speech Emotion Recognition ◽

Deep Convolution Neural Network ◽

Long Short Term Memory ◽

High Level ◽

Better Than

Speech emotion recognition (SER) is a difficult and challenging task because of the affective variances between different speakers. The performances of SER are extremely reliant on the extracted features from speech signals. To establish an effective features extracting and classification model is still a challenging task. In this paper, we propose a new method for SER based on Deep Convolution Neural Network (DCNN) and Bidirectional Long Short-Term Memory with Attention (BLSTMwA) model (DCNN-BLSTMwA). We first preprocess the speech samples by data enhancement and datasets balancing. Secondly, we extract three-channel of log Mel-spectrograms (static, delta, and delta-delta) as DCNN input. Then the DCNN model pre-trained on ImageNet dataset is applied to generate the segment-level features. We stack these features of a sentence into utterance-level features. Next, we adopt BLSTM to learn the high-level emotional features for temporal summarization, followed by an attention layer which can focus on emotionally relevant features. Finally, the learned high-level emotional features are fed into the Deep Neural Network (DNN) to predict the final emotion. Experiments on EMO-DB and IEMOCAP database obtain the unweighted average recall (UAR) of 87.86 and 68.50%, respectively, which are better than most popular SER methods and demonstrate the effectiveness of our propose method.

Download Full-text

Jatropha Curcas Disease Identification With Extreme Learning Machine

Indonesian Journal of Electrical Engineering and Computer Science ◽

10.11591/ijeecs.v12.i2.pp883-888 ◽

2018 ◽

Vol 12 (2) ◽

pp. 883

Author(s):

Triando Hamonangan Saragih ◽

Diny Melsye Nurul Fajri ◽

Wayan Firdaus Mahmudy ◽

Abdul Latief Abadi ◽

Yusuf Priyo Anggodo

Keyword(s):

Neural Network ◽

Expert Systems ◽

Extreme Learning Machine ◽

Jatropha Curcas ◽

Comparison Method ◽

Disease Identification ◽

Maximum Accuracy ◽

Average Accuracy ◽

Learning Machine ◽

Better Than

Jatropha is a plant that has many functions, but this plant can be attacked by various diseases. Expert systems can be applied in identifying so that can help both farmers and extension workers to identify the disease. one of method that can be used is Extreme Learning Machine. Extreme Learning Machine is a method of learning in Neural Network which has a one-time iteration concept in each process. In this study get a maximum accuracy of 66.67% with an average accuracy of 60.61%. This proves the identification using Extreme Learning Machine is better than the comparison method that has been done before.

Download Full-text

EKTRAKSI FITUR MENGGUNAKAN DISCRETE WAVELET TRANSFORM DAN FULL NEIGHBOUR LOCAL BINARY PATTERN UNTUK KLASIFIKASI MAMMOGRAM

Jurnal Ilmiah Teknologi Infomasi Terapan ◽

10.33197/jitter.vol3.iss2.2017.127 ◽

2017 ◽

Vol 3 (2) ◽

Author(s):

Januar Adi Putra ◽

Nanik Suciati ◽

Arya Yudhi Wijaya

Keyword(s):

Neural Network ◽

Wavelet Transform ◽

Discrete Wavelet Transform ◽

Local Binary Pattern ◽

New Method ◽

Discrete Wavelet ◽

Backpropagation Neural Network ◽

Average Accuracy ◽

Mammogram Image ◽

Better Than

[Id]Local binary pattern adalah sebuah kode biner yang menggambarkan pola tekstur lokal. Hal ini dibangun dengan lingkungan batas dengan nilai abu-abu dari pusatnya. Local binary pattern tradisional memiliki beberapa kelemahan yakni varian terhadap rotasi dan pada saat proses thresholding pixel sensitif terhadap noise. Pada penelitian ini diusulkan sebuah metode ektraksi fitur baru untuk mengatasi masalah tersebut, metode tersebut disebut full neighbour local binary pattern (fnlbp). Metode ini nantinya akan dikombinasikan dengan discrete wavelet transform untuk ektraksi fitur dari citra mammogram dengan metode klasifikasi adalah Backpropagation Neural Network (BPNN). Berdasar ujicoba yang telah dilakukan metode usulan mendapatkan rata-rata akurasi yang lebih baik daripada metode local binary pattern tradisional baik yang dikombinasi dengan discrete wavelet transform ataupun tidak. Performa metode usulan full neighbour local binary pattern dapat menghasilkan akurasi yang sempurna yakni 100% baik pada saat menggunakan discrete wavelet transform ataupun tidak, sedangkan akurasi terendah yang didapat adalah 90.49%.Kata Kunci: Ekstraksi fitur, local binary pattern, wavelet, klasifikasi mammogram.[En]Traditional local binary pattern have some disadvantages which is a variant of the rotation and during the thresholding process the pixel is sensitive to noise. At this study the authors proposed a new method of features extraction to solve that problem and this method called full neighbor local binary pattern (fnlbp). This method will be combined with discrete wavelet transform to extract the features of the mammogram image and the classification method is Backpro- pagation Neural Network (BPNN). Based on experiments the result of proposed method in an average accuracy is better than traditional methods of local binary pattern which combined with discrete wavelet transform or not. The performance of the proposed method of full neighbor local binary pattern can produce perfect accuracy that is 100%, this accuracy is reached when using discrete wavelet transform or not, while the lowest accuracy obtained is 90.49%.

Download Full-text