Modified Autoencoder Training and Scoring for Robust Unsupervised Anomaly Detection in Deep Learning

Merrill, Nicholas; Eskandarian, Azim

doi:10.1109/access.2020.2997327

Cited by 33 publications

(5 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…M , N , and C denote the number of pixels along the x and y axis and the number of channels (in this work N = M = 256 and C = 3). The autoencoder training process exploited a dynamic learning rate adjustment and an early stopping function, with a minimum learning rate of 0.0001 and 5 epochs [38]. Once trained, the encoder layer was subsequently used to extract from an RGB input image the 1-D feature vector to be given as input to the feature selection and ripeness classification pipeline.…”

Section: B Ripeness Classification Pipeline Through Rgb Cameramentioning

confidence: 99%

A low-cost multispectral device for in-field fruit ripening assessment

Lauretti,

Tamantini,

Zompanti

et al. 2024

IEEE Sensors J.

View full text Add to dashboard Cite

When assessing fruit ripeness, multispectral sensors offer a cheaper alternative to high-resolution spectroscopy. Additionally, they provide a more robust solution to environmental factors, in contrast to multispectral cameras. However, the low spectral resolution of multispectral devices proposed in the literature may not be sufficient to discriminate with acceptable accuracy a high number of different ripeness classes with machine learning techniques. Moreover, meaningful features to be given as input to the classification models are generally hand-crafted on the basis of the fruit to be evaluated, making the devices not versatile across other fruit categories. This paper aims at overcoming these limitations, by introducing a multispectral device for in-situ fruitripening evaluation that is i) affordable for a wide range of end-users, ii) robust to environmental factors, iii) capable of automatically finding the most meaningful features depending on the fruit categories to evaluate and iv) capable of discriminating a high number of fruit ripeness classes with high accuracy. The device integrates a broadband LED and a VIS-NIR multispectral sensor to gather information about the fruit' ripeness levels through predictive machine learning algorithms. The proposed device was trained on a dataset of 450 spectral measurements acquired from tomatoes at six different ripeness stages. The results demonstrated the high capability of the proposed approach to recognize the tomatoes' ripeness levels (average accuracy of 93.72%).

show abstract

Section: B Ripeness Classification Pipeline Through Rgb Cameramentioning

confidence: 99%

A low-cost multispectral device for in-field fruit ripening assessment

Lauretti,

Tamantini,

Zompanti

et al. 2024

IEEE Sensors J.

View full text Add to dashboard Cite

show abstract

“…Anomaly detection The use of VAEs has been explored for unsupervised anomaly detection in [4], where the model reconstruction error has been used as anomaly score. Our approach differs in that, instead of basing our predictions on the reconstruction error of a single input, which has been shown in [5] to be an unreliable indicator in the unsupervised context, we consider a sequence of input images from the same location. Our problem is better framed as change detection.…”

Section: Application Contextmentioning

confidence: 99%

Unsupervised Change Detection of Extreme Events Using ML On-Board

Růžička¹,

Vaughan²,

Martini³

et al. 2021

Preprint

View full text Add to dashboard Cite

In this paper, we introduce RaVAEn, a lightweight, unsupervised approach for change detection in satellite data based on Variational Auto-Encoders (VAEs) with the specific purpose of on-board deployment. Applications such as disaster management enormously benefit from the rapid availability of satellite observations. Traditionally, data analysis is performed on the ground after all data is transferreddownlinked -to a ground station. Constraint on the downlink capabilities therefore affects any downstream application. In contrast, RaVAEn pre-processes the sampled data directly on the satellite and flags changed areas to prioritise for downlink, shortening the response time. We verified the efficacy of our system on a dataset composed of time series of catastrophic events -which we plan to release alongside this publication -demonstrating that RaVAEn outperforms pixelwise baselines. Finally we tested our approach on resource-limited hardware for assessing computational and memory limitations.

show abstract

“…Today, anomaly detection is broadly used in many research areas such as health monitoring [1], [2], [3], [4], [5], [6] for example heart disease diagnosis [1] and neuromuscular disorders diagnosis [5], environment monitoring such as sewer pipeline fault identification [7] and solar farms anomalies detection [8], and machine condition monitoring [9], [10] for example machinery fault diagnosis [11], [12], [13], [14], [9]. Depending on the anomaly detection problem, it is required to design algorithms which are able to identify anomalies in different types of data such as image [15], [2], [16], [17], video [7], sound signal [9] speech signal [18], sensor signal [19], [5], text [20], spatio-temporal data [4], streaming data [21] and time-series [22], [23]. Hence, it seems that no general solution works for all of the anomaly detection problems.…”

Section: Introductionmentioning

confidence: 99%

Developing a Data-Driven Unsupervised Pattern Recognition Approach for Sensor Signal Anomaly Detection

Iranmehr¹,

Ferreira²,

Böhnert³

et al. 2021

Preprint

View full text Add to dashboard Cite

Coming up with a system for early detection of machine damages and failures is one of the important challenges in the industrial maintenance procedure to avoid additional costs and downtimes. To approach this goal, this paper uses the signal gathered by a sensing system which employed a spintropic sensor to measure the magnetic field around the machine which somehow shows the machine's behaviour. Using this signal and focusing on analysing and processing the signal, this paper develops a data-driven method to recognize signal patterns and subsequently detects anomalies. A challenging task that we succeeded to overcome in this paper is recognizing relevant signal patterns without having any prior knowledge. An algorithm designed for this task is therefore completely unsupervised which makes it consistent and suitable to apply it for the signals gathered for other types of machines. Using both frequency and time domain information, the proposed algorithm, which utilizes signal processing and machine learning techniques, is able to efficiently identify relevant signal patterns. Clustering results on the real data gathered by the aforementioned sensor have shown the high accuracy of 99.38% in recognizing patterns. Furthermore, an anomaly score measure is used and according to its distribution, anomalies are detected appropriately. <br>

show abstract

Modified Autoencoder Training and Scoring for Robust Unsupervised Anomaly Detection in Deep Learning

Cited by 33 publications

References 28 publications

A low-cost multispectral device for in-field fruit ripening assessment

A low-cost multispectral device for in-field fruit ripening assessment

Unsupervised Change Detection of Extreme Events Using ML On-Board

Developing a Data-Driven Unsupervised Pattern Recognition Approach for Sensor Signal Anomaly Detection

Contact Info

Product

Resources

About