Data-Centric Performance Improvement Strategies for Few-Shot Classification of Chemical Sensor Data

2021 ◽  
Vol 10 (1) ◽  
pp. 44
Author(s):  
Bhargavi Mahesh ◽  
Teresa Scholz ◽  
Jana Streit ◽  
Thorsten Graunke ◽  
Sebastian Hettenkofer
Keyword(s):  
Error Analysis ◽  
Chemical Sensor ◽  
Low Cost ◽  
Fine Tuning ◽  
Sensor Data ◽  
Shot Classification ◽  
Average Accuracy ◽  
Voc Mixtures ◽  
Mox Sensors

Metal oxide (MOX) sensors offer a low-cost solution to detect volatile organic compound (VOC) mixtures. However, their operation involves time-consuming heating cycles, leading to a slower data collection and data classification process. This work introduces a few-shot learning approach that promotes rapid classification. In this approach, a model trained on several base classes is fine-tuned to recognize a novel class using a small number (n = 5, 25, 50 and 75) of randomly selected novel class measurements/shots. The used dataset comprises MOX sensor measurements of four different juices (apple, orange, currant and multivitamin) and air, collected over 10-minute phases using a pulse heater signal. While high average accuracy of 82.46 is obtained for five-class classification using 75 shots, the model’s performance depends on the juice type. One-shot validation showed that not all measurements within a phase are representative, necessitating careful shot selection to achieve high classification accuracy. Error analysis revealed contamination of some measurements by the previously measured juice, a characteristic of MOX sensor data that is often overlooked and equivalent to mislabeling. Three strategies are adopted to overcome this: (E1) and (E2) fine-tuning after dropping initial/final measurements and the first half of each phase, respectively, (E3) pretraining with data from the second half of each phase. Results show that each of the strategies performs best for a specific number of shots. E3 results in the highest performance for five-shot learning (accuracy 63.69), whereas E2 yields the best results for 25-/50-shot learning (accuracies 79/87.1) and E1 predicts best for 75-shot learning (accuracy 88.6). Error analysis also showed that, for all strategies, more than 50% of air misclassifications resulted from contamination, but E1 was affected the least. This work demonstrates how strongly data quality can affect prediction performance, especially for few-shot classification methods, and that a data-centric approach can improve the results.

scholarly journals Multi-Class Classification of Lung Diseases Using CNN Models

2021 ◽  
Vol 11 (19) ◽  
pp. 9289
Author(s):  
Min Hong ◽  
Beanbonyka Rim ◽  
Hongchang Lee ◽  
Hyeonung Jang ◽  
Joonho Oh ◽  
...  
Keyword(s):  
Lung Diseases ◽  
Image Data ◽  
National Institutes Of Health ◽  
Fine Tuning ◽  
University Hospital ◽  
Improve Performance ◽  
Average Sensitivity ◽  
Average Accuracy ◽  
Multi Class Classification

In this study, we propose a multi-class classification method by learning lung disease images with Convolutional Neural Network (CNN). As the image data for learning, the U.S. National Institutes of Health (NIH) dataset divided into Normal, Pneumonia, and Pneumothorax and the Cheonan Soonchunhyang University Hospital dataset including Tuberculosis were used. To improve performance, preprocessing was performed with Center Crop while maintaining the aspect ratio of 1:1. As a Noisy Student of EfficientNet B7, fine-tuning learning was performed using the weights learned from ImageNet, and the features of each layer were maximally utilized using the Multi GAP structure. As a result of the experiment, Benchmarks measured with the NIH dataset showed the highest performance among the tested models with an accuracy of 85.32%, and the four-class predictions measured with data from Soonchunhyang University Hospital in Cheonan had an average accuracy of 96.1%, an average sensitivity of 92.2%, an average specificity of 97.4%, and an average inference time of 0.2 s.

scholarly journals Comparative Study of Movie Shot Classification Based on Semantic Segmentation

2020 ◽  
Vol 10 (10) ◽  
pp. 3390
Author(s):  
Hui-Yong Bak ◽  
Seung-Bo Park
Keyword(s):  
Neural Network ◽  
Performance Improvement ◽  
Semantic Segmentation ◽  
Additional Information ◽  
Shot Classification ◽  
Average Accuracy ◽  
The Face ◽  
Shot Type ◽  
Type Classification

The shot-type decision is a very important pre-task in movie analysis due to the vast information, such as the emotion, psychology of the characters, and space information, from the shot type chosen. In order to analyze a variety of movies, a technique that automatically classifies shot types is required. Previous shot type classification studies have classified shot types by the proportion of the face on-screen or using a convolutional neural network (CNN). Studies that have classified shot types by the proportion of the face on-screen have not classified the shot if a person is not on the screen. A CNN classifies shot types even in the absence of a person on the screen, but there are certain shots that cannot be classified because instead of semantically analyzing the image, the method classifies them only by the characteristics and patterns of the image. Therefore, additional information is needed to access the image semantically, which can be done through semantic segmentation. Consequently, in the present study, the performance of shot type classification was improved by preprocessing the semantic segmentation of the frame extracted from the movie. Semantic segmentation approaches the images semantically and distinguishes the boundary relationships among objects. The representative technologies of semantic segmentation include Mask R-CNN and Yolact. A study was conducted to compare and evaluate performance using these as pretreatments for shot type classification. As a result, the average accuracy of shot type classification using a frame preprocessed with semantic segmentation increased by 1.9%, from 93% to 94.9%, when compared with shot type classification using the frame without such preprocessing. In particular, when using ResNet-50 and Yolact, the classification of shot type showed a 3% performance improvement (to 96% accuracy from 93%).

scholarly journals Low-Cost Environmental and Motion Sensor Data for Complex Activity Recognition: Proof of Concept

2020 ◽  
Vol 2 (1) ◽  
pp. 54
Author(s):  
Rok Novak ◽  
David Kocman ◽  
Johanna Amalia Robinson ◽  
Tjaša Kanduč ◽  
Denis Sarigiannis ◽  
...  
Keyword(s):  
Machine Learning ◽  
Low Cost ◽  
Sensor Data ◽  
Motion Sensor ◽  
Data Sets ◽  
Proof Of Concept ◽  
Complex Activity ◽  
Machine Learning Methods ◽  
Complex Activities

The merge of new sensing technologies with machine learning methods can be used as a tool to recognize complex activities. A wearable particulate matter (PM) sensor, in combination with a motion tracker, was provided to 97 individuals for 7 days in two seasons. These data sets were used in three different models, constructed by the classification of activity. Using algorithms IBk, J48 and RandomForest for hourly (minute) values, an accuracy of 31.0 (23.1)%, 28.6 (22.0)% and 35.7 (23.0)%, respectively, was achieved. Most misclassified instances concern vaguely defined activities. Low accuracy can also be explained with the differences in time scales. The accuracy could be improved by more clearly defining the activities and collecting per-minute data.

scholarly journals Deep Belief Network for Spectral–Spatial Classification of Hyperspectral Remote Sensor Data

Sensors ◽  
2019 ◽  
Vol 19 (1) ◽  
pp. 204 ◽  
Author(s):  
Chenming Li ◽  
Yongchang Wang ◽  
Xiaoke Zhang ◽  
Hongmin Gao ◽  
Yao Yang ◽  
...  
Keyword(s):  
Feature Extraction ◽  
Deep Learning ◽  
Optical Sensors ◽  
Spatial Information ◽  
Hyperspectral Image ◽  
Deep Belief Network ◽  
Fine Tuning ◽  
Sensor Data ◽  
Belief Network

With the development of high-resolution optical sensors, the classification of ground objects combined with multivariate optical sensors is a hot topic at present. Deep learning methods, such as convolutional neural networks, are applied to feature extraction and classification. In this work, a novel deep belief network (DBN) hyperspectral image classification method based on multivariate optical sensors and stacked by restricted Boltzmann machines is proposed. We introduced the DBN framework to classify spatial hyperspectral sensor data on the basis of DBN. Then, the improved method (combination of spectral and spatial information) was verified. After unsupervised pretraining and supervised fine-tuning, the DBN model could successfully learn features. Additionally, we added a logistic regression layer that could classify the hyperspectral images. Moreover, the proposed training method, which fuses spectral and spatial information, was tested over the Indian Pines and Pavia University datasets. The advantages of this method over traditional methods are as follows: (1) the network has deep structure and the ability of feature extraction is stronger than traditional classifiers; (2) experimental results indicate that our method outperforms traditional classification and other deep learning approaches.

Actual issues of systematization of scientific knowledge on the conditions of shot in the forensic ballistics

2020 ◽  
Vol 2020 (3) ◽  
pp. 164-169
Author(s):  
Igor' Latyshov ◽  
Fedor Samuylenko
Keyword(s):  
Scientific Knowledge ◽  
Communication Systems ◽  
Production Method ◽  
Natural Phenomenon ◽  
Projectile Energy ◽  
Energy Characteristics ◽  
Shot Classification ◽  
Forensic Ballistics

In this research, there was considered a challenge of constructing a system of scientific knowledge of the shot conditions in judicial ballistics. It was observed that there are underlying factors that are intended to ensureits [scientific knowledge] consistency: identification of the list of shot conditions, which require consideration when solving expert-level research tasks on weapons, cartridges and traces of their action; determination of the communication systems in the course of objects’ interaction, which present the result of exposure to the conditions of the shot; classification of the shot conditions based on the grounds significant for solving scientific and practical problems. The article contains the characteristics of a constructive, functional factor (condition) of weapons and cartridges influence, environmental and fire factors, the structure of the target and its physical properties, situational and spatial factors, and projectile energy characteristics. Highlighted are the forms of connections formed in the course of objects’ interaction, proposed are the author’s classifications of forensically significant shooting conditions with them being divided on the basis of the following criteria: production from the object of interaction, production from a natural phenomenon, production method, results weapon operation and utilization, duration of exposure, type of structural connections between interaction objects, number of conditions that apply when firing and the forming traces.

Sitsen: Passive sitting posture sensing based on wireless devices

2021 ◽  
Vol 17 (7) ◽  
pp. 155014772110248
Author(s):  
Miaoyu Li ◽  
Zhuohan Jiang ◽  
Yutong Liu ◽  
Shuheng Chen ◽  
Marcin Wozniak ◽  
...  
Keyword(s):  
Radio Frequency ◽  
Radio Frequency Identification ◽  
High Performance ◽  
Learning Algorithm ◽  
Low Cost ◽  
Recognition System ◽  
Sitting Posture ◽  
Average Accuracy ◽  
Phase Variations ◽  
Window Approach

Physical health diseases caused by wrong sitting postures are becoming increasingly serious and widespread, especially for sedentary students and workers. Existing video-based approaches and sensor-based approaches can achieve high accuracy, while they have limitations like breaching privacy and relying on specific sensor devices. In this work, we propose Sitsen, a non-contact wireless-based sitting posture recognition system, just using radio frequency signals alone, which neither compromises the privacy nor requires using various specific sensors. We demonstrate that Sitsen can successfully recognize five habitual sitting postures with just one lightweight and low-cost radio frequency identification tag. The intuition is that different postures induce different phase variations. Due to the received phase readings are corrupted by the environmental noise and hardware imperfection, we employ series of signal processing schemes to obtain clean phase readings. Using the sliding window approach to extract effective features of the measured phase sequences and employing an appropriate machine learning algorithm, Sitsen can achieve robust and high performance. Extensive experiments are conducted in an office with 10 volunteers. The result shows that our system can recognize different sitting postures with an average accuracy of 97.02%.

scholarly journals Sparse Autoencoder-based Multi-head Deep Neural Networks for Machinery Fault Diagnostics with Detection of Novelties

2021 ◽  
Vol 34 (1) ◽  
Author(s):  
Zhe Yang ◽  
Dejan Gjorgjevikj ◽  
Jianyu Long ◽  
Yanyang Zi ◽  
Shaohui Zhang ◽  
...  
Keyword(s):  
Fault Diagnosis ◽  
Supervised Classification ◽  
Deep Neural Networks ◽  
Activation Function ◽  
Reconstruction Error ◽  
Fine Tuning ◽  
Fault Diagnostics ◽  
Sparse Autoencoder

AbstractSupervised fault diagnosis typically assumes that all the types of machinery failures are known. However, in practice unknown types of defect, i.e., novelties, may occur, whose detection is a challenging task. In this paper, a novel fault diagnostic method is developed for both diagnostics and detection of novelties. To this end, a sparse autoencoder-based multi-head Deep Neural Network (DNN) is presented to jointly learn a shared encoding representation for both unsupervised reconstruction and supervised classification of the monitoring data. The detection of novelties is based on the reconstruction error. Moreover, the computational burden is reduced by directly training the multi-head DNN with rectified linear unit activation function, instead of performing the pre-training and fine-tuning phases required for classical DNNs. The addressed method is applied to a benchmark bearing case study and to experimental data acquired from a delta 3D printer. The results show that its performance is satisfactory both in detection of novelties and fault diagnosis, outperforming other state-of-the-art methods. This research proposes a novel fault diagnostics method which can not only diagnose the known type of defect, but also detect unknown types of defects.

Sign in / Sign up

Export Citation Format

Share Document