scholarly journals Online Deep Learning Hyperparameter Tuning based on Provenance Analysis

2021 ◽  
Vol 12 (5) ◽  
Author(s):  
Liliane Kunstmann ◽  
Débora Pina ◽  
Filipe Silva ◽  
Aline Paes ◽  
Patrick Valduriez ◽  
...  
Keyword(s):  
Deep Learning ◽  
Data Analysis ◽  
Fine Tuning ◽  
Training Dataset ◽  
Final Decision ◽  
Provenance Analysis ◽  
Online Analysis ◽  
Provenance Data ◽  
Validation Set ◽  
Parameter Values

Training Deep Learning (DL) models require adjusting a series of hyperparameters. Although there are several tools to automatically choose the best hyperparameter configuration, the user is still the main actor to take the final decision. To decide whether the training should continue or try different configurations, the user needs to analyze online the hyperparameters most adequate to the training dataset, observing metrics such as accuracy and loss values. Provenance naturally represents data derivation relationships (i.e., transformations, parameter values, etc.), which provide important support in this data analysis. Most of the existing provenance solutions define their own and proprietary data representations to support DL users in choosing the best hyperparameter configuration, which makes data analysis and interoperability difficult. We present Keras-Prov and its extension, named Keras-Prov++, which provides an analytical dashboard to support online hyperparameter fine-tuning. Different from the current mainstream solutions, Keras-Prov automatically captures the provenance data of DL applications using the W3C PROV recommendation, allowing for hyperparameter online analysis to help the user deciding on changing hyperparameters’ values after observing the performance of the models on a validation set. We provide an experimental evaluation of Keras-Prov++ using AlexNet and a real case study, named DenseED, that acts as a surrogate model for solving equations. During the online analysis, the users identify scenarios that suggest reducing the number of epochs to avoid unnecessary executions and fine-tuning the learning rate to improve the model accuracy.

scholarly journals Covid-19 detection via deep neural network and occlusion sensitivity maps

2021 ◽  
Author(s):  
Noor Ahmad ◽  
Muhammad Aminu ◽  
Mohd Halim Mohd Noor
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Deep Neural Network ◽  
State Of The Art ◽  
Color Images ◽  
Fine Tuning ◽  
Training Dataset ◽  
Learning Approaches ◽  
Learning Models ◽  
Sensitivity Maps

Deep learning approaches have attracted a lot of attention in the automatic detection of Covid-19 and transfer learning is the most common approach. However, majority of the pre-trained models are trained on color images, which can cause inefficiencies when fine-tuning the models on Covid-19 images which are often grayscale. To address this issue, we propose a deep learning architecture called CovidNet which requires a relatively smaller number of parameters. CovidNet accepts grayscale images as inputs and is suitable for training with limited training dataset. Experimental results show that CovidNet outperforms other state-of-the-art deep learning models for Covid-19 detection.

scholarly journals Covid-19 detection via deep neural network and occlusion sensitivity maps

2021 ◽  
Author(s):  
Noor Ahmad ◽  
Muhammad Aminu ◽  
Mohd Halim Mohd Noor
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Deep Neural Network ◽  
State Of The Art ◽  
Color Images ◽  
Fine Tuning ◽  
Training Dataset ◽  
Learning Approaches ◽  
Learning Models ◽  
Sensitivity Maps

Deep learning approaches have attracted a lot of attention in the automatic detection of Covid-19 and transfer learning is the most common approach. However, majority of the pre-trained models are trained on color images, which can cause inefficiencies when fine-tuning the models on Covid-19 images which are often grayscale. To address this issue, we propose a deep learning architecture called CovidNet which requires a relatively smaller number of parameters. CovidNet accepts grayscale images as inputs and is suitable for training with limited training dataset. Experimental results show that CovidNet outperforms other state-of-the-art deep learning models for Covid-19 detection.

scholarly journals DEEP LEARNING BASED ROOF TYPE CLASSIFICATION USING VERY HIGH RESOLUTION AERIAL IMAGERY

2021 ◽  
Vol XLIII-B3-2021 ◽  
pp. 55-60
Author(s):  
M. Buyukdemircioglu ◽  
R. Can ◽  
S. Kocaman
Keyword(s):  
Deep Learning ◽  
High Resolution ◽  
Urban Areas ◽  
Image Features ◽  
Training Data ◽  
Fine Tuning ◽  
Computer Hardware ◽  
Geographical Information ◽  
Training Dataset ◽  
Very High

Abstract. Automatic detection, segmentation and reconstruction of buildings in urban areas from Earth Observation (EO) data are still challenging for many researchers. Roof is one of the most important element in a building model. The three-dimensional geographical information system (3D GIS) applications generally require the roof type and roof geometry for performing various analyses on the models, such as energy efficiency. The conventional segmentation and classification methods are often based on features like corners, edges and line segments. In parallel to the developments in computer hardware and artificial intelligence (AI) methods including deep learning (DL), image features can be extracted automatically. As a DL technique, convolutional neural networks (CNNs) can also be used for image classification tasks, but require large amount of high quality training data for obtaining accurate results. The main aim of this study was to generate a roof type dataset from very high-resolution (10 cm) orthophotos of Cesme, Turkey, and to classify the roof types using a shallow CNN architecture. The training dataset consists 10,000 roof images and their labels. Six roof type classes such as flat, hip, half-hip, gable, pyramid and complex roofs were used for the classification in the study area. The prediction performance of the shallow CNN model used here was compared with the results obtained from the fine-tuning of three well-known pre-trained networks, i.e. VGG-16, EfficientNetB4, ResNet-50. The results show that although our CNN has slightly lower performance expressed with the overall accuracy, it is still acceptable for many applications using sparse data.

scholarly journals A NEW STEREO DENSE MATCHING BENCHMARK DATASET FOR DEEP LEARNING

2021 ◽  
Vol XLIII-B2-2021 ◽  
pp. 405-412
Author(s):  
T. Wu ◽  
B. Vallet ◽  
M. Pierrot-Deseilligny ◽  
E. Rupnik
Keyword(s):  
Deep Learning ◽  
Ground Truth ◽  
Fine Tuning ◽  
Training Dataset ◽  
Ground Truth Data ◽  
Aerial Photogrammetry ◽  
Dense Matching ◽  
Benchmark Datasets ◽  
3D Scene ◽  
The Impact

Abstract. Stereo dense matching is a fundamental task for 3D scene reconstruction. Recently, deep learning based methods have proven effective on some benchmark datasets, for example Middlebury and KITTI stereo. However, it is not easy to find a training dataset for aerial photogrammetry. Generating ground truth data for real scenes is a challenging task. In the photogrammetry community, many evaluation methods use digital surface models (DSM) to generate the ground truth disparity for the stereo pairs, but in this case interpolation may bring errors in the estimated disparity. In this paper, we publish a stereo dense matching dataset based on ISPRS Vaihingen dataset, and use it to evaluate some traditional and deep learning based methods. The evaluation shows that learning-based methods outperform traditional methods significantly when the fine tuning is done on a similar landscape. The benchmark also investigates the impact of the base to height ratio on the performance of the evaluated methods. The dataset can be found in https://github.com/whuwuteng/benchmark_ISPRS2021.

scholarly journals Improving Performance of Seismic Fault Detection by Fine-Tuning the Convolutional Neural Network Pre-Trained with Synthetic Samples

Energies ◽  
2021 ◽  
Vol 14 (12) ◽  
pp. 3650
Author(s):  
Zhe Yan ◽  
Zheng Zhang ◽  
Shaoyong Liu
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Fault Detection ◽  
Seismic Data ◽  
Learning Strategy ◽  
Real Data ◽  
Fine Tuning ◽  
Detection Methods ◽  
Training Dataset ◽  
Detection Accuracy

Fault interpretation is an important part of seismic structural interpretation and reservoir characterization. In the conventional approach, faults are detected as reflection discontinuity or abruption and are manually tracked in post-stack seismic data, which is time-consuming. In order to improve efficiency, a variety of automatic fault detection methods have been proposed, among which widespread attention has been given to deep learning-based methods. However, deep learning techniques require a large amount of marked seismic samples as a training dataset. Although the amount of synthetic seismic data can be guaranteed and the labels are accurate, the difference between synthetic data and real data still exists. To overcome this drawback, we apply a transfer learning strategy to improve the performance of automatic fault detection by deep learning methods. We first pre-train a deep neural network with synthetic seismic data. Then we retrain the network with real seismic samples. We use a random sample consensus (RANSAC) method to obtain real seismic samples and generate corresponding labels automatically. Three real 3D examples are included to demonstrate that the fault detection accuracy of the pre-trained network models can be greatly improved by retraining the network with a few amount of real seismic samples.

scholarly journals Large-Scale Printed Chinese Character Recognition for ID Cards Using Deep Learning and Few Samples Transfer Learning

2021 ◽  
Author(s):  
Yi-Quan Li ◽  
Hao-Sen Chang ◽  
Daw-Tung Lin
Keyword(s):  
Neural Networks ◽  
Deep Learning ◽  
Transfer Learning ◽  
Character Recognition ◽  
Large Scale ◽  
Data Augmentation ◽  
Training Data ◽  
Fine Tuning ◽  
Training Dataset ◽  
Model Parameters

In the field of computer vision, large-scale image classification tasks are both important and highly challenging. With the ongoing advances in deep learning and optical character recognition (OCR) technologies, neural networks designed to perform large-scale classification play an essential role in facilitating OCR systems. In this study, we developed an automatic OCR system designed to identify up to 13,070 large-scale printed Chinese characters by using deep learning neural networks and fine-tuning techniques. The proposed framework comprises four components, including training dataset synthesis and background simulation, image preprocessing and data augmentation, the process of training the model, and transfer learning. The training data synthesis procedure is composed of a character font generation step and a background simulation process. Three background models are proposed to simulate the factors of the background noise and anti-counterfeiting patterns on ID cards. To expand the diversity of the synthesized training dataset, rotation and zooming data augmentation are applied. A massive dataset comprising more than 19.6 million images was thus created to accommodate the variations in the input images and improve the learning capacity of the CNN model. Subsequently, we modified the GoogLeNet neural architecture by replacing the FC layer with a global average pooling layer to avoid overfitting caused by a massive amount of training data. Consequently, the number of model parameters was reduced. Finally, we employed the transfer learning technique to further refine the CNN model using a small number of real data samples. Experimental results show that the overall recognition performance of the proposed approach is significantly better than that of prior methods and thus demonstrate the effectiveness of proposed framework, which exhibited a recognition accuracy as high as 99.39% on the constructed real ID card dataset.

scholarly journals Label3DMaize: toolkit for 3D point cloud data annotation of maize shoots

GigaScience ◽  
2021 ◽  
Vol 10 (5) ◽  
Author(s):  
Teng Miao ◽  
Weiliang Wen ◽  
Yinglun Li ◽  
Sheng Wu ◽  
Chao Zhu ◽  
...  
Keyword(s):  
Deep Learning ◽  
Point Cloud ◽  
Training Dataset ◽  
Growth Stages ◽  
3D Point Cloud ◽  
Plant Structure ◽  
Data Annotation ◽  
Cloud Data ◽  
Point Cloud Segmentation ◽  
Different Growth Stages

Abstract Background The 3D point cloud is the most direct and effective data form for studying plant structure and morphology. In point cloud studies, the point cloud segmentation of individual plants to organs directly determines the accuracy of organ-level phenotype estimation and the reliability of the 3D plant reconstruction. However, highly accurate, automatic, and robust point cloud segmentation approaches for plants are unavailable. Thus, the high-throughput segmentation of many shoots is challenging. Although deep learning can feasibly solve this issue, software tools for 3D point cloud annotation to construct the training dataset are lacking. Results We propose a top-to-down point cloud segmentation algorithm using optimal transportation distance for maize shoots. We apply our point cloud annotation toolkit for maize shoots, Label3DMaize, to achieve semi-automatic point cloud segmentation and annotation of maize shoots at different growth stages, through a series of operations, including stem segmentation, coarse segmentation, fine segmentation, and sample-based segmentation. The toolkit takes ∼4–10 minutes to segment a maize shoot and consumes 10–20% of the total time if only coarse segmentation is required. Fine segmentation is more detailed than coarse segmentation, especially at the organ connection regions. The accuracy of coarse segmentation can reach 97.2% that of fine segmentation. Conclusion Label3DMaize integrates point cloud segmentation algorithms and manual interactive operations, realizing semi-automatic point cloud segmentation of maize shoots at different growth stages. The toolkit provides a practical data annotation tool for further online segmentation research based on deep learning and is expected to promote automatic point cloud processing of various plants.

scholarly journals Study on Radar Echo-Filling in an Occlusion Area by a Deep Learning Algorithm

2021 ◽  
Vol 13 (9) ◽  
pp. 1779
Author(s):  
Xiaoyan Yin ◽  
Zhiqun Hu ◽  
Jiafeng Zheng ◽  
Boyong Li ◽  
Yuanyuan Zuo
Keyword(s):  
Deep Learning ◽  
Loss Function ◽  
Learning Algorithm ◽  
Weather Radar ◽  
Loss Functions ◽  
Training Dataset ◽  
Echo Intensity ◽  
Common Mean ◽  
Deep Learning Algorithm ◽  
Radar Beam

Radar beam blockage is an important error source that affects the quality of weather radar data. An echo-filling network (EFnet) is proposed based on a deep learning algorithm to correct the echo intensity under the occlusion area in the Nanjing S-band new-generation weather radar (CINRAD/SA). The training dataset is constructed by the labels, which are the echo intensity at the 0.5° elevation in the unblocked area, and by the input features, which are the intensity in the cube including multiple elevations and gates corresponding to the location of bottom labels. Two loss functions are applied to compile the network: one is the common mean square error (MSE), and the other is a self-defined loss function that increases the weight of strong echoes. Considering that the radar beam broadens with distance and height, the 0.5° elevation scan is divided into six range bands every 25 km to train different models. The models are evaluated by three indicators: explained variance (EVar), mean absolute error (MAE), and correlation coefficient (CC). Two cases are demonstrated to compare the effect of the echo-filling model by different loss functions. The results suggest that EFnet can effectively correct the echo reflectivity and improve the data quality in the occlusion area, and there are better results for strong echoes when the self-defined loss function is used.

Sign in / Sign up

Export Citation Format

Share Document