Search results for: deep learning image model

Authors: Mahasa Zahirnia

Abstract:

The purpose of this paper is to explore new and effective ways of optimizing satellite images using artificial intelligence, and the process of implementing reinforcement learning to enhance the quality of data captured within the image. In our implementation of Bellman's Reinforcement Learning equations, associated state diagrams, and multi-stage image processing, we were able to enhance image quality, detect and define objects. Reinforcement learning is the differentiator in the area of artificial intelligence, and Q-Learning relies on trial and error to achieve its goals. The reward system that is embedded in Q-Learning allows the agent to self-evaluate its performance and decide on the best possible course of action based on the current and future environment. Results show that within a simulated environment, built on the images that are commercially available, the rate of detection was 40-90%. Reinforcement learning through Q-Learning algorithm is not just desired but required design criteria for image optimization and enhancements. The proposed methods presented are a cost effective method of resolving uncertainty of the data because reinforcement learning finds ideal policies to manage the process using a smaller sample of images.

Keywords: Q-learning, image optimization, reinforcement learning, Markov decision process

Procedia PDF Downloads 192

Authors: Sophia Shi

Abstract:

Acute kidney injury (AKI) is the sudden onset of kidney damage in which the kidneys cannot filter waste from the blood, requiring emergency hospitalization. AKI patient mortality rate is high in the ICU and is virtually impossible for doctors to predict because it is so unexpected. Currently, there is no hybrid model predicting AKI that takes advantage of two types of data. De-identified patient data from the MIMIC-III database and de-identified kidney images and corresponding patient records from the Beijing Hospital of the Ministry of Health were collected. Using data features including serum creatinine among others, two numeric models using MIMIC and Beijing Hospital data were built, and with the hospital ultrasounds, an image-only model was built. Convolutional neural networks (CNN) were used, VGG and Resnet for numeric data and Resnet for image data, and they were combined into a hybrid model by concatenating feature maps of both types of models to create a new input. This input enters another CNN block and then two fully connected layers, ending in a binary output after running through Softmax and additional code. The hybrid model successfully predicted AKI and the highest AUROC of the model was 0.953, achieving an accuracy of 90% and F1-score of 0.91. This model can be implemented into urgent clinical settings such as the ICU and aid doctors by assessing the risk of AKI shortly after the patient’s admission to the ICU, so that doctors can take preventative measures and diminish mortality risks and severe kidney damage.

Keywords: Acute kidney injury, Convolutional neural network, Hybrid deep learning, Patient record data, ResNet, Ultrasound kidney images, VGG

Procedia PDF Downloads 112

Authors: Sai Maniveer Adapa, Sai Guptha Perla, Adithya Reddy P.

Abstract:

Thyroid nodules can often be diagnosed with ultrasound imaging, although differentiating between benign and malignant nodules can be challenging for medical professionals. This work suggests a novel approach to increase the precision of thyroid nodule identification by combining machine learning and deep learning. The new approach first extracts information from the ultrasound pictures using a deep learning method known as a convolutional autoencoder. A support vector machine, a type of machine learning model, is then trained using these features. With an accuracy of 92.52%, the support vector machine can differentiate between benign and malignant nodules. This innovative technique may decrease the need for pointless biopsies and increase the accuracy of thyroid nodule detection.

Keywords: thyroid tumor diagnosis, ultrasound images, deep learning, machine learning, convolutional auto-encoder, support vector machine

Procedia PDF Downloads 32

Authors: Rui Mao, Heming Ji, Xiaoyu Wang

Abstract:

Keystroke dynamics are widely used in identity recognition. It has the advantage that the individual typing rhythm is difficult to imitate. It also supports continuous authentication through the keyboard without extra devices. The existing keystroke dynamics authentication methods based on machine learning have a drawback in supporting relatively complex scenarios with massive data. There are drawbacks to both feature extraction and model optimization in these methods. To overcome the above weakness, an authentication model of keystroke dynamics based on deep learning is proposed. The model uses feature vectors formed by keystroke content and keystroke time. It ensures efficient continuous authentication by cooperating attention mechanisms with the combination of CNN and Bi-LSTM. The model has been tested with Open Data Buffalo dataset, and the result shows that the FRR is 3.09%, FAR is 3.03%, and EER is 4.23%. This proves that the model is efficient and accurate on continuous authentication.

Keywords: keystroke dynamics, identity authentication, deep learning, CNN, LSTM

Procedia PDF Downloads 130

Authors: Abe D. Desta

Abstract:

This study suggests a deep learning-based method for identifying and following moving objects in surveillance video. The proposed method uses a fast regional convolution neural network (F-RCNN) trained on a substantial dataset of vehicle images to first detect vehicles. A Kalman filter and a data association technique based on a Hungarian algorithm are then used to monitor the observed vehicles throughout time. However, in general, F-RCNN algorithms have been shown to be effective in achieving high detection accuracy and robustness in this research study. For example, in one study The study has shown that the vehicle detection and tracking, the system was able to achieve an accuracy of 97.4%. In this study, the F-RCNN algorithm was compared to other popular object detection algorithms and was found to outperform them in terms of both detection accuracy and speed. The presented system, which has application potential in actual surveillance systems, shows the usefulness of deep learning approaches in vehicle detection and tracking.

Keywords: artificial intelligence, computer vision, deep learning, fast-regional convolutional neural networks, feature extraction, vehicle tracking

Procedia PDF Downloads 85

Authors: Zabeehullah, Fahim Arif, Yawar Abbas

Abstract:

Software Defined Network (SDN) is a next genera-tion networking model which simplifies the traditional network complexities and improve the utilization of constrained resources. Currently, most of the SDN based Internet of Things(IoT) environments use traditional network routing strategies which work on the basis of max or min metric value. However, IoT network heterogeneity, dynamic traffic flow and complexity demands intelligent and self-adaptive routing algorithms because traditional routing algorithms lack the self-adaptions, intelligence and efficient utilization of resources. To some extent, SDN, due its flexibility, and centralized control has managed the IoT complexity and heterogeneity but still Software Defined IoT (SDIoT) lacks intelligence. To address this challenge, we proposed a model called Deep Routing Strategy (DRS) which uses Deep Learning algorithm to perform routing in SDIoT intelligently and efficiently. Our model uses real-time traffic for training and learning. Results demonstrate that proposed model has achieved high accuracy and low packet loss rate during path selection. Proposed model has also outperformed benchmark routing algorithm (OSPF). Moreover, proposed model provided encouraging results during high dynamic traffic flow.

Keywords: SDN, IoT, DL, ML, DRS

Procedia PDF Downloads 91

Authors: Liang Wang, Beste F. Yuksel, David Guy Brizan

Abstract:

A computational model of affect that can distinguish between spontaneous and posed smiles with no errors on a large, popular data set using deep learning techniques is presented in this paper. A Long Short-Term Memory (LSTM) classifier, a type of Recurrent Neural Network, is utilized and compared to human classification. Results showed that while human classification (mean of 0.7133) was above chance, the LSTM model was more accurate than human classification and other comparable state-of-the-art systems. Additionally, a high accuracy rate was maintained with small amounts of training videos (70 instances). The derivation of important features to further understand the success of our computational model were analyzed, and it was inferred that thousands of pairs of points within the eyes and mouth are important throughout all time segments in a smile. This suggests that distinguishing between a posed and spontaneous smile is a complex task, one which may account for the difficulty and lower accuracy of human classification compared to machine learning models.

Keywords: affective computing, affect detection, computer vision, deep learning, human-computer interaction, machine learning, posed smile detection, spontaneous smile detection

Procedia PDF Downloads 108

Authors: Kashika P. H.

Abstract:

The development of computer vision systems has been greatly aided by the efficient and precise detection of images and videos. Although the ability to recognize and comprehend images is a strength of the human brain, employing technology to tackle this issue is exceedingly challenging. In the past few years, the use of Deep Learning algorithms to treat object detection has dramatically expanded. One of the key issues in the realm of image recognition is the recognition and detection of certain notable people from randomly acquired photographs. Face recognition uses a way to identify, assess, and compare faces for a variety of purposes, including user identification, user counting, and classification. With the aid of an accessible deep learning-based API, this article intends to recognize various faces of people and their facial descriptors more accurately. The purpose of this study is to locate suitable individuals and deliver accurate information about them by using the Amazon Rekognition system to identify a specific human from a vast image dataset. We have chosen the Amazon Rekognition system, which allows for more accurate face analysis, face comparison, and face search, to tackle this difficulty.

Keywords: Amazon rekognition, API, deep learning, computer vision, face detection, text detection

Procedia PDF Downloads 87

Authors: Smitha Haridasan, Zelalem Demissie, Atri Dutta, Ajita Rattani

Abstract:

Aided by the wind, all it takes is one ember and a few minutes to create a wildfire. Wildfires are growing in frequency and size due to climate change. Wildfires and its consequences are one of the major environmental concerns. Every year, millions of hectares of forests are destroyed over the world, causing mass destruction and human casualties. Thus early detection of wildfire becomes a critical component to mitigate this threat. Many computer vision-based techniques have been proposed for the early detection of forest fire using video surveillance. Several computer vision-based methods have been proposed to predict and detect forest fires at various spectrums, namely, RGB, HSV, and YCbCr. The aim of this paper is to propose a multi-spectral deep learning model that combines information from different spectrums at intermediate layers for accurate fire detection. A heterogeneous dataset assembled from publicly available datasets is used for model training and evaluation in this study. The experimental results show that multi-spectral deep learning models could obtain an improvement of about 4.68 % over those based on a single spectrum for fire detection.

Keywords: deep learning, forest fire detection, multi-spectral learning, natural hazard detection

Procedia PDF Downloads 211

Authors: Romisaa Ali

Abstract:

The Policy Gradient approach is one of the deep reinforcement learning families that combines deep neural networks (DNN) with reinforcement learning RL to discover the optimum of the control problem through experience gained from the interaction between the robot and its surroundings. In contrast to earlier policy gradient algorithms, which were unable to handle these two types of error because of over-or under-estimation introduced by the deep neural network model, this article will discuss the state-of-the-art SOTA policy gradient technique, trust region policy optimization (TRPO), by applying this method in various environments compared to another policy gradient method, the Proximal Policy Optimization (PPO), to explain their robust optimization, using this SOTA to gather experience data during various training phases after observing the impact of hyper-parameters on neural network performance.

Keywords: deep neural networks, deep reinforcement learning, proximal policy optimization, state-of-the-art, trust region policy optimization

Procedia PDF Downloads 145

Authors: Simone Leonardi, Luca Ardito

Abstract:

Micro-influencers are effective elements in the marketing strategies of companies and institutions because of their capability to create an hyper-engaged audience around a specific topic of interest. In recent years, many scientific approaches and commercial tools have handled the task of detecting this type of social media users. These strategies adopt solutions ranging from rule based machine learning models to deep neural networks and graph analysis on text, images, and account information. This work compares the existing solutions and proposes an ensemble method to generalize them with different input data and social media platforms. The deployed solution combines deep learning models on unstructured data with statistical machine learning models on structured data. We retrieve both social media accounts information and multimedia posts on Twitter and Instagram. These data are mapped into feature vectors for an eXtreme Gradient Boosting (XGBoost) classifier. Sixty different topics have been analyzed to build a rule based gold standard dataset and to compare the performances of our approach against baseline classifiers. We prove the effectiveness of our work by comparing the accuracy, precision, recall, and f1 score of our model with different configurations and architectures. We obtained an accuracy of 0.91 with our best performing model.

Keywords: deep learning, gradient boosting, image processing, micro-influencers, NLP, social media

Procedia PDF Downloads 154

Authors: Ons Slimene, Aroua Taamallah, Maha Khemaja

Abstract:

In recent years, human activity recognition (HAR) has been a key area of research due to its diverse applications. It has garnered increasing attention in the field of computer vision. HAR plays an important role in people’s daily lives as it has the ability to learn advanced knowledge about human activities from data. In HAR, activities are usually represented by exploiting different types of sensors, such as embedded sensors or visual sensors. However, these sensors have limitations, such as local obstacles, image-related obstacles, sensor unreliability, and consumer concerns. Recently, several deep learning-based approaches have been proposed for HAR and these approaches are classified into two categories based on the type of data used: vision-based approaches and sensor-based approaches. This research paper highlights the importance of multimodal data fusion from skeleton data obtained from videos and data generated by embedded sensors using deep neural networks for achieving HAR. We propose a deep multimodal fusion network based on a twostream architecture. These two streams use the Convolutional Neural Network combined with the Bidirectional LSTM (CNN BILSTM) to process skeleton data and data generated by embedded sensors and the fusion at the feature level is considered. The proposed model was evaluated on a public OPPORTUNITY++ dataset and produced a accuracy of 96.77%.

Keywords: human activity recognition, action recognition, sensors, vision, human-centric sensing, deep learning, context-awareness

Procedia PDF Downloads 75

Authors: Susan Diamond

Abstract:

Machine Learning workloads have traditionally been run in high-performance computing (HPC) environments, where users log in to dedicated machines and utilize the attached GPUs to run training jobs on huge datasets. Training of large neural network models is very resource intensive, and even after exploiting parallelism and accelerators such as GPUs, a single training job can still take days. Consequently, the cost of hardware is a barrier to entry. Even when upfront cost is not a concern, the lead time to set up such an HPC environment takes months from acquiring hardware to set up the hardware with the right set of firmware, software installed and configured. Furthermore, scalability is hard to achieve in a rigid traditional lab environment. Therefore, it is slow to react to the dynamic change in the artificial intelligent industry. Watson Deep Learning as a service, a cloud-based deep learning platform that mitigates the long lead time and high upfront investment in hardware. It enables robust and scalable sharing of resources among the teams in an organization. It is designed for on-demand cloud environments. Providing a similar user experience in a multi-tenant cloud environment comes with its own unique challenges regarding fault tolerance, performance, and security. Watson Deep Learning as a service tackles these challenges and present a deep learning stack for the cloud environments in a secure, scalable and fault-tolerant manner. It supports a wide range of deep-learning frameworks such as Tensorflow, PyTorch, Caffe, Torch, Theano, and MXNet etc. These frameworks reduce the effort and skillset required to design, train, and use deep learning models. Deep Learning as a service is used at IBM by AI researchers in areas including machine translation, computer vision, and healthcare. 

Keywords: deep learning, machine learning, cognitive computing, model training

Procedia PDF Downloads 188

Authors: Yuanchao Sang, Zhihao Gong, Longsheng Chen, Long Chen

Abstract:

Clothing fashion represents a human’s aesthetic appreciation towards everyday outfits and appetite for fashion, and it reflects the development of status in society, humanity, and economics. However, modelling fashion by machine is extremely challenging because fashion is too abstract to be efficiently described by machines. Even human beings can hardly reach a consensus about fashion. In this paper, we are dedicated to answering a fundamental fashion-related problem: what image feature best describes clothing fashion? To address this issue, we have designed and evaluated various image features, ranging from traditional low-level hand-crafted features to mid-level style awareness features to various current popular deep neural network-based features, which have shown state-of-the-art performance in various vision tasks. In summary, we tested the following 9 feature representations: color, texture, shape, style, convolutional neural networks (CNNs), CNNs with distance metric learning (CNNs&DML), AutoEncoder, CNNs with multiple layer combination (CNNs&MLC) and CNNs with dynamic feature clustering (CNNs&DFC). Finally, we validated the performance of these features on two publicly available datasets. Quantitative and qualitative experimental results on both intra-domain and inter-domain fashion clothing image retrieval showed that deep learning based feature representations far outweigh traditional hand-crafted feature representation. Additionally, among all deep learning based methods, CNNs with explicit feature clustering performs best, which shows feature clustering is essential for discriminative fashion feature representation.

Keywords: convolutional neural network, feature representation, image processing, machine modelling

Procedia PDF Downloads 118

Authors: Adam Beaunoyer, Cory Beaunoyer, Mohammed Elmorsy, Hanan Saleh

Abstract:

This research aims to create a reinforcement learning agent capable of racing in challenging simulated environments with a low collision count. We present a reinforcement learning agent that can navigate challenging tracks using both a Deep Q-Network (DQN) and a Soft Actor-Critic (SAC) method. A challenging track includes curves, jumps, and varying road widths throughout. Using open-source code on Github, the environment used in this research is based on the 1995 racing game WipeOut. The proposed reinforcement learning agent can navigate challenging tracks rapidly while maintaining low racing completion time and collision count. The results show that the SAC model outperforms the DQN model by a large margin. We also propose an alternative multiple-car model that can navigate the track without colliding with other vehicles on the track. The SAC model is the basis for the multiple-car model, where it can complete the laps quicker than the single-car model but has a higher collision rate with the track wall.

Keywords: reinforcement learning, soft actor-critic, deep q-network, self-driving cars, artificial intelligence, gaming

Procedia PDF Downloads 18

Authors: Luther Ollier, Sylvie Thiria, Anastase Charantonis, Carlos E. Mejia, Michel Crépon

Abstract:

Sea Surface Height Anomaly (SLA) is a signature of the sub-mesoscale dynamics of the upper ocean. Sea Surface Temperature (SST) is driven by these dynamics and can be used to improve the spatial interpolation of SLA fields. In this study, we focused on the temporal evolution of SLA fields. We explored the capacity of deep learning (DL) methods to predict short-term SLA fields using SST fields. We used simulated daily SLA and SST data from the Mercator Global Analysis and Forecasting System, with a resolution of (1/12)◦ in the North Atlantic Ocean (26.5-44.42◦N, -64.25–41.83◦E), covering the period from 1993 to 2019. Using a slightly modified image-to-image convolutional DL architecture, we demonstrated that SST is a relevant variable for controlling the SLA prediction. With a learning process inspired by the teaching-forcing method, we managed to improve the SLA forecast at five days by using the SST fields as additional information. We obtained predictions of a 12 cm (20 cm) error of SLA evolution for scales smaller than mesoscales and at time scales of 5 days (20 days), respectively. Moreover, the information provided by the SST allows us to limit the SLA error to 16 cm at 20 days when learning the trajectory.

Keywords: deep-learning, altimetry, sea surface temperature, forecast

Procedia PDF Downloads 68

Authors: Jinil Patel, Sarthak Patel, Sarthak Thakkar, Deepti Saraswat

Abstract:

Recently, the COVID-19 outbreak has spread across the world, leading the World Health Organization to classify it as a global pandemic. To save the patient’s life, the COVID-19 symptoms have to be identified. But using an AI (Artificial Intelligence) model to identify COVID-19 symptoms within the allotted time was challenging. The RT-PCR test was found to be inadequate in determining the COVID status of a patient. To determine if the patient has COVID-19 or not, a Computed Tomography Scan (CT scan) of patient is a better alternative. It will be challenging to compile and store all the data from various hospitals on the server, though. Federated learning, therefore, aids in resolving this problem. Certain deep learning models help to classify Covid-19. This paper will have detailed work of certain deep learning models like VGG19, ResNet50, MobileNEtv2, and Deep Learning Aggregation (DLA) along with maintaining privacy with encryption.

Keywords: federated learning, COVID-19, CT-scan, homomorphic encryption, ResNet50, VGG-19, MobileNetv2, DLA

Procedia PDF Downloads 50

Authors: Soofiyan Atar, Adil Shaikh, Sahil Rajpurkar, Pragnesh Bhalala, Aniket Desai, Irfan Siddavatam

Abstract:

Deep reinforcement learning (deep RL) algorithms leverage the symbolic power of complex controllers by automating it by mapping sensory inputs to low-level actions. Deep RL eliminates the complex robot dynamics with minimal engineering. Deep RL provides high-risk involvement by directly implementing it in real-world scenarios and also high sensitivity towards hyperparameters. Tuning of hyperparameters on a pneumatic quadruped robot becomes very expensive through trial-and-error learning. This paper presents an automated learning control for a pneumatic quadruped robot using sample efficient deep Q learning, enabling minimal tuning and very few trials to learn the neural network. Long training hours may degrade the pneumatic cylinder due to jerk actions originated through stochastic weights. We applied this method to the pneumatic quadruped robot, which resulted in a hopping gait. In our process, we eliminated the use of a simulator and acquired a stable gait. This approach evolves so that the resultant gait matures more sturdy towards any stochastic changes in the environment. We further show that our algorithm performed very well as compared to programmed gait using robot dynamics.

Keywords: model-based reinforcement learning, gait stability, supervised learning, pneumatic quadruped

Procedia PDF Downloads 289

Authors: Endrick Barnacin, Jean-Luc Henry, Jimmy Nagau, Jack Molinie

Abstract:

Palynology is a field of interest in many disciplines due to its multiple applications: chronological dating, climatology, allergy treatment, and honey characterization. Unfortunately, the analysis of a pollen slide is a complicated and time consuming task that requires the intervention of experts in the field, which are becoming increasingly rare due to economic and social conditions. That is why the need for automation of this task is urgent. A lot of studies have investigated the subject using different standard image processing descriptors and sometimes hand-crafted ones.In this work, we make a comparative study between classical feature extraction methods (Shape, GLCM, LBP, and others) and Deep Learning (CNN, Autoencoders, Transfer Learning) to perform a recognition task over 80 regional pollen species. It has been found that the use of Transfer Learning seems to be more precise than the other approaches

Keywords: pollens identification, features extraction, pollens classification, automated palynology

Procedia PDF Downloads 114

Authors: Jimeng Shi, Mahek Jain, Giri Narasimhan

Abstract:

Time Series Forecasting (TSF) is used to predict the target variables at a future time point based on the learning from previous time points. To keep the problem tractable, learning methods use data from a fixed-length window in the past as an explicit input. In this paper, we study how the performance of predictive models changes as a function of different look-back window sizes and different amounts of time to predict the future. We also consider the performance of the recent attention-based Transformer models, which have had good success in the image processing and natural language processing domains. In all, we compare four different deep learning methods (RNN, LSTM, GRU, and Transformer) along with a baseline method. The dataset (hourly) we used is the Beijing Air Quality Dataset from the UCI website, which includes a multivariate time series of many factors measured on an hourly basis for a period of 5 years (2010-14). For each model, we also report on the relationship between the performance and the look-back window sizes and the number of predicted time points into the future. Our experiments suggest that Transformer models have the best performance with the lowest Mean Average Errors (MAE = 14.599, 23.273) and Root Mean Square Errors (RSME = 23.573, 38.131) for most of our single-step and multi-steps predictions. The best size for the look-back window to predict 1 hour into the future appears to be one day, while 2 or 4 days perform the best to predict 3 hours into the future.

Keywords: air quality prediction, deep learning algorithms, time series forecasting, look-back window

Procedia PDF Downloads 138

Authors: Nidhal K. Azawi, John M. Gauch

Abstract:

Colorectal cancer is one of the leading causes of cancer death in the US and the world, which is why millions of colonoscopy examinations are performed annually. Unfortunately, noise, specular highlights, and motion artifacts corrupt many images in a typical colonoscopy exam. The goal of our research is to produce automated techniques to detect and correct or remove these noninformative images from colonoscopy videos, so physicians can focus their attention on informative images. In this research, we first automatically extract features from images. Then we use machine learning and deep neural network to classify colonoscopy images as either informative or noninformative. Our results show that we achieve image classification accuracy between 92-98%. We also show how the removal of noninformative images together with image alignment can aid in the creation of image panoramas and other visualizations of colonoscopy images.

Keywords: colonoscopy classification, feature extraction, image alignment, machine learning

Procedia PDF Downloads 234

Authors: Heewon Jeong, Seongpyo Kim, Joon Ha Kim

Abstract:

Over the last few tens of years, the coastal regions of Korea have experienced red tide algal blooms, which are harmful and toxic to both humans and marine organisms due to their potential threat. It was accelerated owing to eutrophication by human activities, certain oceanic processes, and climate change. Previous studies have tried to monitoring and predicting the algae concentration of the ocean with the bio-optical algorithms applied to color images of the satellite. However, the accurate estimation of algal blooms remains problems to challenges because of the complexity of coastal waters. Therefore, this study suggests a new method to identify the concentration of red tide algal bloom from images of geostationary ocean color imager (GOCI) which are representing the water environment of the sea in Korea. The method employed GOCI images, which took the water leaving radiances centered at 443nm, 490nm and 660nm respectively, as well as observed weather data (i.e., humidity, temperature and atmospheric pressure) for the database to apply optical characteristics of algae and train deep learning algorithm. Convolution neural network (CNN) was used to extract the significant features from the images. And then artificial neural network (ANN) was used to estimate the concentration of algae from the extracted features. For training of the deep learning model, backpropagation learning strategy is developed. The established methods were tested and compared with the performances of GOCI data processing system (GDPS), which is based on standard image processing algorithms and optical algorithms. The model had better performance to estimate algae concentration than the GDPS which is impossible to estimate greater than 5mg/m³. Thus, deep learning model trained successfully to assess algae concentration in spite of the complexity of water environment. Furthermore, the results of this system and methodology can be used to improve the performances of remote sensing. Acknowledgement: This work was supported by the 'Climate Technology Development and Application' research project (#K07731) through a grant provided by GIST in 2017.

Keywords: deep learning, algae concentration, remote sensing, satellite

Procedia PDF Downloads 167

Authors: Azeddine El-Hassouny, Chandrashekhar Meshram, Geraldin Nanfack

Abstract:

In recent years, learning from data with noisy labels (Label Noise) has been a major concern in supervised learning. This problem has become even more worrying in Deep Learning, where the generalization capabilities have been questioned lately. Indeed, deep learning requires a large amount of data that is generally collected by search engines, which frequently return data with unreliable labels. In this paper, we investigate the Label Noise in Deep Learning using variational inference. Our contributions are : (1) exploiting Label Noise concept where the true labels are learnt using reparameterization variational inference, while observed labels are learnt discriminatively. (2) the noise transition matrix is learnt during the training without any particular process, neither heuristic nor preliminary phases. The theoretical results shows how true label distribution can be learned by variational inference in any discriminate neural network, and the effectiveness of our approach is proved in several target datasets, such as MNIST and CIFAR32.

Keywords: label noise, deep learning, discrete latent variable, variational inference, MNIST, CIFAR32

Procedia PDF Downloads 100

Authors: John Pomerat, Aviv Segev

Abstract:

Recently, deep learning has had many theoretical breakthroughs. For deep learning to be successful in the industry, however, there need to be practical algorithms capable of handling many real-world hiccups preventing the immediate application of a learning algorithm. Although AI promises to revolutionize the healthcare industry, getting access to patient data in order to train learning algorithms has not been easy. One proposed solution to this is data- sharing. In this paper, we propose an alternative protocol, based on multi-party computation, to train deep learning models while maintaining both the privacy and security of training data. We examine three methods of training neural networks in this way: Transfer learning, average ensemble learning, and series network learning. We compare these methods to the equivalent model obtained through data-sharing across two different experiments. Additionally, we address the security concerns of this protocol. While the motivating example is healthcare, our findings regarding multi-party computation of neural network training are purely theoretical and have use-cases outside the domain of healthcare.

Keywords: neural network aggregation, multi-party computation, transfer learning, average ensemble learning

Procedia PDF Downloads 138

Authors: Jaiden X. Schraut

Abstract:

Medical Imaging Segmentation of Chest X-rays is used for the purpose of identification and differentiation of lung cancer, pneumonia, COVID-19, and similar respiratory diseases. Widespread application of computer-supported perception methods into the diagnostic pipeline has been demonstrated to increase prognostic accuracy and aid doctors in efficiently treating patients. Modern models attempt the task of segmentation and classification separately and improve diagnostic efficiency; however, to further enhance this process, this paper proposes a multi-output network that follows a U-Net architecture for image segmentation output and features an additional CNN module for auxiliary classification output. The proposed model achieves a final Jaccard Index of .9634 for image segmentation and a final accuracy of .9600 for classification on the COVID-19 radiography database.

Keywords: chest X-ray, deep learning, image segmentation, image classification

Procedia PDF Downloads 113

Authors: Selvavinayagan Babiharan

Abstract:

This research aims to develop a deep learning platform to enhance mathematics education for secondary students in Sri Lanka. The platform will be designed to incorporate interactive and user-friendly features to engage students in active learning and promote their mathematical skills. The proposed platform will be developed using TensorFlow and Keras, two widely used deep learning frameworks. The system will be trained on a large dataset of math problems, which will be collected from Sri Lankan school curricula. The results of this research will contribute to the improvement of mathematics education in Sri Lanka and provide a valuable tool for teachers to enhance the learning experience of their students.

Keywords: information technology, education, machine learning, mathematics

Procedia PDF Downloads 60

Authors: Harold Mauricio Díaz-Vargas, Cristian Alfonso Jimenez-Castaño, David Augusto Cárdenas-Peña, Guillermo Alberto Ortiz-Gómez, Alvaro Angel Orozco-Gutierrez

Abstract:

Precise identification of the nerves is a crucial task performed by anesthesiologists for an effective Peripheral Nerve Blocking (PNB). Now, anesthesiologists use ultrasound imaging equipment to guide the PNB and detect nervous structures. However, visual identification of the nerves from ultrasound images is difficult, even for trained specialists, due to artifacts and low contrast. The recent advances in deep learning make neural networks a potential tool for accurate nerve segmentation systems, so addressing the above issues from raw data. The most widely spread U-Net network yields pixel-by-pixel segmentation by encoding the input image and decoding the attained feature vector into a semantic image. This work proposes a conditioning approach and encoder pre-training to enhance the nerve segmentation of traditional U-Nets. Conditioning is achieved by the one-hot encoding of the kind of target nerve a the network input, while the pre-training considers five well-known deep networks for image classification. The proposed approach is tested in a collection of 619 US images, where the best C-UNet architecture yields an 81% Dice coefficient, outperforming the 74% of the best traditional U-Net. Results prove that pre-trained models with the conditional approach outperform their equivalent baseline by supporting learning new features and enriching the discriminant capability of the tested networks.

Keywords: nerve segmentation, U-Net, deep learning, ultrasound imaging, peripheral nerve blocking

Procedia PDF Downloads 81

Authors: Wei Li, Tuo Yang

Abstract:

In order to satisfy the real-time and accurate requirements of cigarette smoke detection in complex scenes, a cigarette smoke detection technology based on the combination of deep learning and color features was proposed. Firstly, based on the color features of cigarette smoke, the suspicious cigarette smoke area in the image is extracted. Secondly, combined with the efficiency of cigarette smoke detection and the problem of network overfitting, a network model for cigarette smoke detection was designed according to YOLOV3 algorithm to reduce the false detection rate. The experimental results show that the method is feasible and effective, and the accuracy of cigarette smoke detection is up to 99.13%, which satisfies the requirements of real-time cigarette smoke detection in complex scenes.

Keywords: deep learning, computer vision, cigarette smoke detection, YOLOV3, color feature extraction

Procedia PDF Downloads 64

Authors: K. Bresilla, L. Manfrini, B. Morandi, A. Boini, G. Perulli, L. C. Grappadelli

Abstract:

Image/video processing for fruit in the tree using hard-coded feature extraction algorithms have shown high accuracy during recent years. While accurate, these approaches even with high-end hardware are computationally intensive and too slow for real-time systems. This paper details the use of deep convolution neural networks (CNNs), specifically an algorithm (YOLO - You Only Look Once) with 24+2 convolution layers. Using deep-learning techniques eliminated the need for hard-code specific features for specific fruit shapes, color and/or other attributes. This CNN is trained on more than 5000 images of apple and pear fruits on 960 cores GPU (Graphical Processing Unit). Testing set showed an accuracy of 90%. After this, trained data were transferred to an embedded device (Raspberry Pi gen.3) with camera for more portability. Based on correlation between number of visible fruits or detected fruits on one frame and the real number of fruits on one tree, a model was created to accommodate this error rate. Speed of processing and detection of the whole platform was higher than 40 frames per second. This speed is fast enough for any grasping/harvesting robotic arm or other real-time applications.

Keywords: artificial intelligence, computer vision, deep learning, fruit recognition, harvesting robot, precision agriculture

Procedia PDF Downloads 392

Authors: Guanghua Zhang, Fubao Wang, Weijun Duan

Abstract:

Convolution neural network (CNN) has attracted more and more attention on recent years. Especially in the field of computer vision and image classification. However, unsupervised learning with CNN has received less attention than supervised learning. In this work, we use a new powerful tool which is deep convolutional generative adversarial networks (DCGANs) to generate images from Sloan Digital Sky Survey. Training by various star and galaxy images, it shows that both the generator and the discriminator are good for unsupervised learning. In this paper, we also took several experiments to choose the best value for hyper-parameters and which could help to stabilize the training process and promise a good quality of the output.

Keywords: convolution neural network, discriminator, generator, unsupervised learning

Procedia PDF Downloads 242