scholarly journals Litter Detection with Deep Learning: A Comparative Study

Sensors ◽  
2022 ◽  
Vol 22 (2) ◽  
pp. 548
Author(s):  
Manuel Córdova ◽  
Allan Pinto ◽  
Christina Carrozzo Hellevik ◽  
Saleh Abdel-Afou Alaliyat ◽  
Ibrahim A. Hameed ◽  
...  

Pollution in the form of litter in the natural environment is one of the great challenges of our times. Automated litter detection can help assess waste occurrences in the environment. Different machine learning solutions have been explored to develop litter detection tools, thereby supporting research, citizen science, and volunteer clean-up initiatives. However, to the best of our knowledge, no work has investigated the performance of state-of-the-art deep learning object detection approaches in the context of litter detection. In particular, no studies have focused on the assessment of those methods aiming their use in devices with low processing capabilities, e.g., mobile phones, typically employed in citizen science activities. In this paper, we fill this literature gap. We performed a comparative study involving state-of-the-art CNN architectures (e.g., Faster RCNN, Mask-RCNN, EfficientDet, RetinaNet and YOLO-v5), two litter image datasets and a smartphone. We also introduce a new dataset for litter detection, named PlastOPol, composed of 2418 images and 5300 annotations. The experimental results demonstrate that object detectors based on the YOLO family are promising for the construction of litter detection solutions, with superior performance in terms of detection accuracy, processing time, and memory footprint.

2020 ◽  
Vol 12 (3) ◽  
pp. 582 ◽  
Author(s):  
Rui Li ◽  
Shunyi Zheng ◽  
Chenxi Duan ◽  
Yang Yang ◽  
Xiqi Wang

In recent years, researchers have paid increasing attention on hyperspectral image (HSI) classification using deep learning methods. To improve the accuracy and reduce the training samples, we propose a double-branch dual-attention mechanism network (DBDA) for HSI classification in this paper. Two branches are designed in DBDA to capture plenty of spectral and spatial features contained in HSI. Furthermore, a channel attention block and a spatial attention block are applied to these two branches respectively, which enables DBDA to refine and optimize the extracted feature maps. A series of experiments on four hyperspectral datasets show that the proposed framework has superior performance to the state-of-the-art algorithm, especially when the training samples are signally lacking.


Sensors ◽  
2019 ◽  
Vol 19 (5) ◽  
pp. 1058 ◽  
Author(s):  
Yang-Yang Zheng ◽  
Jian-Lei Kong ◽  
Xue-Bo Jin ◽  
Xiao-Yi Wang ◽  
Min Zuo

Intelligence has been considered as the major challenge in promoting economic potential and production efficiency of precision agriculture. In order to apply advanced deep-learning technology to complete various agricultural tasks in online and offline ways, a large number of crop vision datasets with domain-specific annotation are urgently needed. To encourage further progress in challenging realistic agricultural conditions, we present the CropDeep species classification and detection dataset, consisting of 31,147 images with over 49,000 annotated instances from 31 different classes. In contrast to existing vision datasets, images were collected with different cameras and equipment in greenhouses, captured in a wide variety of situations. It features visually similar species and periodic changes with more representative annotations, which have supported a stronger benchmark for deep-learning-based classification and detection. To further verify the application prospect, we provide extensive baseline experiments using state-of-the-art deep-learning classification and detection models. Results show that current deep-learning-based methods achieve well performance in classification accuracy over 99%. While current deep-learning methods achieve only 92% detection accuracy, illustrating the difficulty of the dataset and improvement room of state-of-the-art deep-learning models when applied to crops production and management. Specifically, we suggest that the YOLOv3 network has good potential application in agricultural detection tasks.


2019 ◽  
Author(s):  
Michael Uhl ◽  
Van Dinh Tran ◽  
Rolf Backofen

AbstractCLIP-seq is the state-of-the-art technique to experimentally determine transcriptome-wide binding sites of RNA-binding proteins (RBPs). However, it relies on gene expression which can be highly variable between conditions, and thus cannot provide a complete picture of the RBP binding landscape. This necessitates the use of computational methods to predict missing binding sites. Here we present GraphProt2, a computational RBP binding site prediction method based on graph convolutional neural networks (GCN). In contrast to current CNN methods, GraphProt2 supports variable length input as well as the possibility to accurately predict nucleotide-wise binding profiles. We demonstrate its superior performance compared to GraphProt and a CNN-based method on single as well as combined CLIP-seq datasets.


Author(s):  
Ioannis Prapas ◽  
Behrouz Derakhshan ◽  
Alireza Rezaei Mahdiraji ◽  
Volker Markl

AbstractDeep Learning (DL) has consistently surpassed other Machine Learning methods and achieved state-of-the-art performance in multiple cases. Several modern applications like financial and recommender systems require models that are constantly updated with fresh data. The prominent approach for keeping a DL model fresh is to trigger full retraining from scratch when enough new data are available. However, retraining large and complex DL models is time-consuming and compute-intensive. This makes full retraining costly, wasteful, and slow. In this paper, we present an approach to continuously train and deploy DL models. First, we enable continuous training through proactive training that combines samples of historical data with new streaming data. Second, we enable continuous deployment through gradient sparsification that allows us to send a small percentage of the model updates per training iteration. Our experimental results with LeNet5 on MNIST and modern DL models on CIFAR-10 show that proactive training keeps models fresh with comparable—if not superior—performance to full retraining at a fraction of the time. Combined with gradient sparsification, sparse proactive training enables very fast updates of a deployed model with arbitrarily large sparsity, reducing communication per iteration up to four orders of magnitude, with minimal—if any—losses in model quality. Sparse training, however, comes at a price; it incurs overhead on the training that depends on the size of the model and increases the training time by factors ranging from 1.25 to 3 in our experiments. Arguably, a small price to pay for successfully enabling the continuous training and deployment of large DL models.


Sensors ◽  
2019 ◽  
Vol 19 (14) ◽  
pp. 3166 ◽  
Author(s):  
Cao ◽  
Song ◽  
Song ◽  
Xiao ◽  
Peng

Lane detection is an important foundation in the development of intelligent vehicles. To address problems such as low detection accuracy of traditional methods and poor real-time performance of deep learning-based methodologies, a lane detection algorithm for intelligent vehicles in complex road conditions and dynamic environments was proposed. Firstly, converting the distorted image and using the superposition threshold algorithm for edge detection, an aerial view of the lane was obtained via region of interest extraction and inverse perspective transformation. Secondly, the random sample consensus algorithm was adopted to fit the curves of lane lines based on the third-order B-spline curve model, and fitting evaluation and curvature radius calculation were then carried out on the curve. Lastly, by using the road driving video under complex road conditions and the Tusimple dataset, simulation test experiments for lane detection algorithm were performed. The experimental results show that the average detection accuracy based on road driving video reached 98.49%, and the average processing time reached 21.5 ms. The average detection accuracy based on the Tusimple dataset reached 98.42%, and the average processing time reached 22.2 ms. Compared with traditional methods and deep learning-based methodologies, this lane detection algorithm had excellent accuracy and real-time performance, a high detection efficiency and a strong anti-interference ability. The accurate recognition rate and average processing time were significantly improved. The proposed algorithm is crucial in promoting the technological level of intelligent vehicle driving assistance and conducive to the further improvement of the driving safety of intelligent vehicles.


2018 ◽  
Vol 9 (24) ◽  
pp. 5441-5451 ◽  
Author(s):  
Andreas Mayr ◽  
Günter Klambauer ◽  
Thomas Unterthiner ◽  
Marvin Steijaert ◽  
Jörg K. Wegner ◽  
...  

The to date largest comparative study of nine state-of-the-art drug target prediction methods finds that deep learning outperforms all other competitors. The results are based on a benchmark of 1300 assays and half a million compounds.


Cells ◽  
2020 ◽  
Vol 9 (8) ◽  
pp. 1756 ◽  
Author(s):  
Abdul Wahab ◽  
Omid Mahmoudi ◽  
Jeehong Kim ◽  
Kil To Chong

N4-methylcytosine as one kind of modification of DNA has a critical role which alters genetic performance such as protein interactions, conformation, stability in DNA as well as the regulation of gene expression same cell developmental and genomic imprinting. Some different 4mC site identifiers have been proposed for various species. Herein, we proposed a computational model, DNC4mC-Deep, including six encoding techniques plus a deep learning model to predict 4mC sites in the genome of F. vesca, R. chinensis, and Cross-species dataset. It was demonstrated by the 10-fold cross-validation test to get superior performance. The DNC4mC-Deep obtained 0.829 and 0.929 of MCC on F. vesca and R. chinensis training dataset, respectively, and 0.814 on cross-species. This means the proposed method outperforms the state-of-the-art predictors at least 0.284 and 0.265 on F. vesca and R. chinensis training dataset in turn. Furthermore, the DNC4mC-Deep achieved 0.635 and 0.565 of MCC on F. vesca and R. chinensis independent dataset, respectively, and 0.562 on cross-species which shows it can achieve the best performance to predict 4mC sites as compared to the state-of-the-art predictor.


2020 ◽  
Vol 12 (1) ◽  
pp. 90-108
Author(s):  
Mahmoud Kalash ◽  
Mrigank Rochan ◽  
Noman Mohammed ◽  
Neil Bruce ◽  
Yang Wang ◽  
...  

In this article, the authors propose a deep learning framework for malware classification. There has been a huge increase in the volume of malware in recent years which poses serious security threats to financial institutions, businesses, and individuals. In order to combat the proliferation of malware, new strategies are essential to quickly identify and classify malware samples. Nowadays, machine learning approaches are becoming popular for malware classification. However, most of these approaches are based on shallow learning algorithms (e.g. SVM). Recently, convolutional neural networks (CNNs), a deep learning approach, have shown superior performance compared to traditional learning algorithms, especially in tasks such as image classification. Inspired by this, the authors propose a CNN-based architecture to classify malware samples. They convert malware binaries to grayscale images and subsequently train a CNN for classification. Experiments on two challenging malware classification datasets, namely Malimg and Microsoft, demonstrate that their method outperforms competing state-of-the-art algorithms.


2021 ◽  
Vol 7 (11) ◽  
pp. 223
Author(s):  
Gabriele Antonio De Vitis ◽  
Antonio Di Tecco ◽  
Pierfrancesco Foglia ◽  
Cosimo Antonio Prete

During the production of pharmaceutical glass tubes, a machine-vision based inspection system can be utilized to perform the high-quality check required by the process. The necessity to improve detection accuracy, and increase production speed determines the need for fast solutions for defects detection. Solutions proposed in literature cannot be efficiently exploited due to specific factors that characterize the production process. In this work, we have derived an algorithm that does not change the detection quality compared to state-of-the-art proposals, but does determine a drastic reduction in the processing time. The algorithm utilizes an adaptive threshold based on the Sigma Rule to detect blobs, and applies a threshold to the variation of luminous intensity along a row to detect air lines. These solutions limit the detection effects due to the tube’s curvature, and rotation and vibration of the tube, which characterize glass tube production. The algorithm has been compared with state-of-the-art solutions. The results demonstrate that, with the algorithm proposed, the processing time of the detection phase is reduced by 86%, with an increase in throughput of 268%, achieving greater accuracy in detection. Performance is further improved by adopting Region of Interest reduction techniques. Moreover, we have developed a tuning procedure to determine the algorithm’s parameters in the production batch change. We assessed the performance of the algorithm in a real environment using the “certification” functionality of the machine. Furthermore, we observed that out of 1000 discarded tubes, nine should not have been discarded and a further seven should have been discarded.


2016 ◽  
Author(s):  
Feng Liu ◽  
Hao Li ◽  
Chao Ren ◽  
Xiaochen Bo ◽  
Wenjie Shu

AbstractTranscriptional enhancers are non-coding segments of DNA that play a central role in the spatiotemporal regulation of gene expression programs. However, systematically and precisely predicting enhancers remain a major challenge. Although existing methods have achieved some success in enhancer prediction, they still suffer from many issues. We developed a deep learning-based algorithmic framework named PEDLA (https://github.com/wenjiegroup/PEDLA), which can directly learn an enhancer predictor from massively heterogeneous data and generalize in ways that are mostly consistent across various cell types/tissues. We first trained PEDLA with 1,114-dimensional heterogeneous features in H1 cells, and we demonstrated that our PEDLA framework integrates diverse heterogeneous features and gives state-of-the-art performance relative to five existing methods for enhancer prediction. We further extended PEDLA to iteratively learn from 22 training cell types/tissues. Our results showed that PEDLA manifested superior performance consistency in both training and independent test sets. On average, PEDLA achieved 95.0% accuracy and a 96.8% geometric mean (GM) across 22 training cell types/tissues, as well as 95.7% accuracy and a 96.8% GM across 20 independent test cell types/tissues. Together, our work illustrates the power of harnessing state-of-the-art deep learning techniques to consistently identify regulatory elements at a genome-wide scale from massively heterogeneous data across diverse cell types/tissues.


Sign in / Sign up

Export Citation Format

Share Document