Performance Of Deep Learning Models Research Articles

A Deep Learning Approach for Malnutrition Detection

The timely detection of malnutrition in children is of paramount importance, as it allows for early intervention and treatment. This proactive approach not only prevents further health deterioration but also fosters proper growth, minimizing the long-term consequences of malnutrition, such as stunted growth, impaired cognitive development, and increased vulnerability to diseases. Our work encompasses the creation of a new dataset comprising images of children in Healthy, Undernourished, Stunting, and Wasting categories. The core objective is to assess the deep learning model performance in classifying these children images. The experimentation is carried out by varying epochs, batch size, optimizers AdamW, Adamax, and RMSprop; and different values of the learning rate 0.1, 0.01, 0.001, and 0.0001 during model training. The model is trained on image dataset constructed by cleaning images generated by the stable diffusion model. The model is tested on randomly selected child images from websites. The model successfully classified two classes with 95% accuracy, 97.6% F1 score, precision 97.6%, and 97.6% recall with Adam optimizers, 0.0001 learning rate, and Batch size 4. Additionally, for the four-class categorization scenario, the study broadens the classification. The model achieved 88.87% accuracy, 90.3% recall, 90.2% precision, and an F1 score of 90% for four-class categorization with AdamW optimization, 0.0001 learning rate, and batch size 6. These results are satisfactory for prediction of malnutrition category in children.

Utilizing Polarization Diversity in GBSAR Data-Based Object Classification.

In recent years, the development of intelligent sensor systems has experienced remarkable growth, particularly in the domain of microwave and millimeter wave sensing, thanks to the increased availability of affordable hardware components. With the development of smart Ground-Based Synthetic Aperture Radar (GBSAR) system called GBSAR-Pi, we previously explored object classification applications based on raw radar data. Building upon this foundation, in this study, we analyze the potential of utilizing polarization information to improve the performance of deep learning models based on raw GBSAR data. The data are obtained with a GBSAR operating at 24 GHz with both vertical (VV) and horizontal (HH) polarization, resulting in two matrices (VV and HH) per observed scene. We present several approaches demonstrating the integration of such data into classification models based on a modified ResNet18 architecture. We also introduce a novel Siamese architecture tailored to accommodate the dual input radar data. The results indicate that a simple concatenation method is the most promising approach and underscore the importance of considering antenna polarization and merging strategies in deep learning applications based on radar data.

Synthesizing 3D Multi-Contrast Brain Tumor MRIs Using Tumor Mask Conditioning.

Data scarcity and data imbalance are two major challenges in training deep learning models on medical images, such as brain tumor MRI data. The recent advancements in generative artificial intelligence have opened new possibilities for synthetically generating MRI data, including brain tumor MRI scans. This approach can be a potential solution to mitigate the data scarcity problem and enhance training data availability. This work focused on adapting the 2D latent diffusion models to generate 3D multi-contrast brain tumor MRI data with a tumor mask as the condition. The framework comprises two components: a 3D autoencoder model for perceptual compression and a conditional 3D Diffusion Probabilistic Model (DPM) for generating high-quality and diverse multi-contrast brain tumor MRI samples, guided by a conditional tumor mask. Unlike existing works that focused on generating either 2D multi-contrast or 3D single-contrast MRI samples, our models generate multi-contrast 3D MRI samples. We also integrated a conditional module within the UNet backbone of the DPM to capture the semantic class-dependent data distribution driven by the provided tumor mask to generate MRI brain tumor samples based on a specific brain tumor mask. We trained our models using two brain tumor datasets: The Cancer Genome Atlas (TCGA) public dataset and an internal dataset from the University of Texas Southwestern Medical Center (UTSW). The models were able to generate high-quality 3D multi-contrast brain tumor MRI samples with the tumor location aligned by the input condition mask. The quality of the generated images was evaluated using the Fréchet Inception Distance (FID) score. This work has the potential to mitigate the scarcity of brain tumor data and improve the performance of deep learning models involving brain tumor MRI data.

Improving accuracy of vascular access quality classification in hemodialysis patients using deep learning with K highest score feature selection.

To develop and evaluate a novel feature selection technique, using photoplethysmography (PPG) sensors, for enhancing the performance of deep learning models in classifying vascular access quality in hemodialysis patients. This cross-sectional study involved creating a novel feature selection method based on SelectKBest principles, specifically designed to optimize deep learning models for PPG sensor data, in hemodialysis patients. The method effectiveness was assessed by comparing the performance of multiple deep learning models using the feature selection approach versus complete feature set. The model with the highest accuracy was then trained and tested using a 70:30 approach, respectively, with the full dataset and the SelectKBest dataset. Performance results were compared using Student's paired t-test. Data from 398 hemodialysis patients were included. The 1-dimensional convolutional neural network (CNN1D) displayed the highest accuracy among different models. Implementation of the SelectKBest-based feature selection technique resulted in a statistically significant improvement in the CNN1D model's performance, achieving an accuracy of 92.05% (with feature selection) versus 90.79% (with full feature set). These findings suggest that the newly developed feature selection approach might aid in accurately predicting vascular access quality in hemodialysis patients. This advancement may contribute to the development of reliable diagnostic tools for identifying vascular complications, such as stenosis, potentially improving patient outcomes and their quality of life.

Convolutional neural network deep learning model accurately detects rectal cancer in endoanal ultrasounds

BackgroundImaging is vital for assessing rectal cancer, with endoanal ultrasound (EAUS) being highly accurate in large tertiary medical centers. However, EAUS accuracy drops outside such settings, possibly due to varied examiner experience and fewer examinations. This underscores the need for an AI-based system to enhance accuracy in non-specialized centers. This study aimed to develop and validate deep learning (DL) models to differentiate rectal cancer in standard EAUS images.MethodsA transfer learning approach with fine-tuned DL architectures was employed, utilizing a dataset of 294 images. The performance of DL models was assessed through a tenfold cross-validation.ResultsThe DL diagnostics model exhibited a sensitivity and accuracy of 0.78 each. In the identification phase, the automatic diagnostic platform achieved an area under the curve performance of 0.85 for diagnosing rectal cancer.ConclusionsThis research demonstrates the potential of DL models in enhancing rectal cancer detection during EAUS, especially in settings with lower examiner experience. The achieved sensitivity and accuracy suggest the viability of incorporating AI support for improved diagnostic outcomes in non-specialized medical centers.

Implications of Deep Learning for Stock Market Forecasting

This research explores the effectiveness of using deep learning in predicting stock market movements. This research uses rigorous methods to bring out the performance of deep learning models, compare them with traditional methods, and identify critical factors that influence stock market predictions. The research results show that deep learning models, especially LSTM and CNN-LSTM architectures, can achieve satisfactory levels of accuracy and outperform traditional methods by capturing patterns in complex stock market data. In addition, this research identifies external and internal factors that influence predictions of stock market movements. This research's practical and theoretical implications highlight the potential of deep learning in improving investment decision-making and understanding financial market dynamics. Recommendations for future research include exploration of advanced deep learning techniques, integration with traditional methods, emphasis on risk management strategies, continuous evaluation of model performance, and provision of training and education to encourage analysts and investors to adopt this technology. By implementing these recommendations, the potential of deep learning models in financial analysis can be optimized, ultimately improving market efficiency and investment returns.

Improving spatial transferability of deep learning models for small-field crop yield prediction

Predicting crop yield using deep learning (DL) and remote sensing is a promising technique in agriculture. In smallholder agriculture (<2 ha), where 84% of the farms operate globally, it is crucial to build a model that can be useful across several fields (high spatial transferability). However, enhancing spatial model transferability in a small-scale setting faces significant challenges, including spatial autocorrelation, heterogeneity and scale dependence of spatial dynamics, as well as the need to address limited data points. This study aimed to test the hypothesis that spatial cross validation (SCV) is a more suitable model validation practice than random cross validation (RCV) to enhance model transferability for spatial prediction in a small-scale farming setting. We compared the performances of DL models that predict crop yield for several settings including three crop types and two DL architectures based on RCV with and without overlapping samples and SCV. Notably, we conducted model performance tests on external, equally sized fields instead of the field used for training. We used high resolution RGB imagery taken with a drone as input. Our results show that the models using SCV outperformed those using RCV when the models were tested on external fields (on average r = 0.37 for SCV, r = 0.18 for RCV with overlap and r = 0.07 without), even though the models using SCV showed a substantially lower performance for cross validation (CV) than those using RCV (r with SCV and RCV w/o overlap = 0.73 and 0.98/0.73, respectively). The results suggest that RCV leads to over-optimism by overfitting the spatial structure and remembering image-specific information (so called memorization). Our study offers the first empirical evidence in agriculture that SCV is preferable to RCV in small field settings for making DL models more transferable.

Enhancing Federated Learning With Server-Side Unlabeled Data by Adaptive Client and Data Selection

Federated learning (FL) has been widely applied to collaboratively train deep learning (DL) models on massive end devices (i.e., clients). Due to the limited storage capacity and high labeling cost, the data on each client may be insufficient for model training. Conversely, in cloud datacenters, there exist large-scale unlabeled data, which are easy to collect from public access (e.g., social media). Herein, we propose the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Ada-FedSemi</i> system, which leverages both on-device labeled data and in-cloud unlabeled data to boost the performance of DL models. In each round, local models are aggregated to produce pseudo-labels for the unlabeled data, which are utilized to enhance the global model. Considering that the number of participating clients and the quality of pseudo-labels will have a significant impact on the training performance, we introduce a multi-armed bandit (MAB) based online algorithm to adaptively determine the participating fraction and confidence threshold. Besides, to alleviate the impact of stragglers, we assign local models of different depths for heterogeneous clients. Extensive experiments on benchmark models and datasets show that given the same resource budget, the model trained by Ada-FedSemi achieves 3% <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\sim$</tex-math></inline-formula> 14.8% higher test accuracy than that of the baseline methods. When achieving the same test accuracy, Ada-FedSemi saves up to 48% training cost, compared with the baselines. Under the scenario with heterogeneous clients, the proposed HeteroAda-FedSemi can further speed up the training process by 1.3× <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\sim 1.5\times$</tex-math></inline-formula> .

A systematic analyses of different bioinformatics pipelines for genomic data and its impact on deep learning models for chromatin loop prediction.

Genomic data analysis has witnessed a surge in complexity and volume, primarily driven by the advent of high-throughput technologies. In particular, studying chromatin loops and structures has become pivotal in understanding gene regulation and genome organization. This systematic investigation explores the realm of specialized bioinformatics pipelines designed specifically for the analysis of chromatin loops and structures. Our investigation incorporates two protein (CTCF and Cohesin) factor-specific loop interaction datasets from six distinct pipelines, amassing a comprehensive collection of 36 diverse datasets. Through a meticulous review of existing literature, we offer a holistic perspective on the methodologies, tools and algorithms underpinning the analysis of this multifaceted genomic feature. We illuminate the vast array of approaches deployed, encompassing pivotal aspects such as data preparation pipeline, preprocessing, statistical features and modelling techniques. Beyond this, we rigorously assess the strengths and limitations inherent in these bioinformatics pipelines, shedding light on the interplay between data quality and the performance of deep learning models, ultimately advancing our comprehension of genomic intricacies.

Improving the performance of supervised deep learning for regulatory genomics using phylogenetic augmentation.

Supervised deep learning is used to model the complex relationship between genomic sequence and regulatory function. Understanding how these models make predictions can provide biological insight into regulatory functions. Given the complexity of the sequence to regulatory function mapping (the cis-regulatory code), it has been suggested that the genome contains insufficient sequence variation to train models with suitable complexity. Data augmentation is a widely used approach to increase the data variation available for model training, however current data augmentation methods for genomic sequence data are limited. Inspired by the success of comparative genomics, we show that augmenting genomic sequences with evolutionarily related sequences from other species, which we term phylogenetic augmentation, improves the performance of deep learning models trained on regulatory genomic sequences to predict high-throughput functional assay measurements. Additionally, we show that phylogenetic augmentation can rescue model performance when the training set is down-sampled and permits deep learning on a real-world small dataset, demonstrating that this approach improves data efficiency. Overall, this data augmentation method represents a solution for improving model performance that is applicable to many supervised deep-learning problems in genomics. The open-source GitHub repository agduncan94/phylogenetic_augmentation_paper includes the code for rerunning the analyses here and recreating the figures.

Detection and Classification of Hysteroscopic Images Using Deep Learning.

Although hysteroscopy with endometrial biopsy is the gold standard in the diagnosis of endometrial pathology, the gynecologist experience is crucial for a correct diagnosis. Deep learning (DL), as an artificial intelligence method, might help to overcome this limitation. Unfortunately, only preliminary findings are available, with the absence of studies evaluating the performance of DL models in identifying intrauterine lesions and the possible aid related to the inclusion of clinical factors in the model. To develop a DL model as an automated tool for detecting and classifying endometrial pathologies from hysteroscopic images. A monocentric observational retrospective cohort study was performed by reviewing clinical records, electronic databases, and stored videos of hysteroscopies from consecutive patients with pathologically confirmed intrauterine lesions at our Center from January 2021 to May 2021. Retrieved hysteroscopic images were used to build a DL model for the classification and identification of intracavitary uterine lesions with or without the aid of clinical factors. Study outcomes were DL model diagnostic metrics in the classification and identification of intracavitary uterine lesions with and without the aid of clinical factors. We reviewed 1500 images from 266 patients: 186 patients had benign focal lesions, 25 benign diffuse lesions, and 55 preneoplastic/neoplastic lesions. For both the classification and identification tasks, the best performance was achieved with the aid of clinical factors, with an overall precision of 80.11%, recall of 80.11%, specificity of 90.06%, F1 score of 80.11%, and accuracy of 86.74 for the classification task, and overall detection of 85.82%, precision of 93.12%, recall of 91.63%, and an F1 score of 92.37% for the identification task. Our DL model achieved a low diagnostic performance in the detection and classification of intracavitary uterine lesions from hysteroscopic images. Although the best diagnostic performance was obtained with the aid of clinical data, such an improvement was slight.

Exploring the potential of data augmentation in poetry generation with small-scale corpora

Poetry generation is a complex task in the field of natural language processing, especially when working with small datasets. Data augmentation techniques have been shown to be an effective way to improve the performance of deep learning models in various tasks, including image classification and speech recognition. Therefore, this study focuses on exploring the impact of four different data augmentation methods - Synonym Replacement, Random Insertion, Random Swap, and Random Deletion - on the performance of poetry generation with a small poetry dataset. The results of the study reveal that Random Insertion performed well in terms of Bilingual Evaluation Understudy (BLEU), Recall-Oriented Understudy for Gisting Evaluation (ROUGE), and manual evaluation when compared to other data augmentation techniques. Synonym Replacement performed poorly in all three evaluations. This study confirms the potential value of data augmentation technology in poetry generation tasks and provides innovative perspectives and directions for future research in this area. Data augmentation can be employed to help address the problem of limited data in poetry generation tasks and enhance the efficiency of deep learning models. Future research could focus on exploring more advanced data augmentation techniques and their impact on poetry generation tasks.

Enhancing Global Estimation of Fine Particulate Matter Concentrations by Including Geophysical a Priori Information in Deep Learning.

Global fine particulate matter (PM2.5) assessment is impeded by a paucity of monitors. We improve estimation of the global distribution of PM2.5 concentrations by developing, optimizing, and applying a convolutional neural network with information from satellite-, simulation-, and monitor-based sources to predict the local bias in monthly geophysical a priori PM2.5 concentrations over 1998-2019. We develop a loss function that incorporates geophysical a priori estimates and apply it in model training to address the unrealistic results produced by mean-square-error loss functions in regions with few monitors. We introduce novel spatial cross-validation for air quality to examine the importance of considering spatial properties. We address the sharp decline in deep learning model performance in regions distant from monitors by incorporating the geophysical a priori PM2.5. The resultant monthly PM2.5 estimates are highly consistent with spatial cross-validation PM2.5 concentrations from monitors globally and regionally. We withheld 10% to 99% of monitors for testing to evaluate the sensitivity and robustness of model performance to the density of ground-based monitors. The model incorporating the geophysical a priori PM2.5 concentrations remains highly consistent with observations globally even under extreme conditions (e.g., 1% for training, R2 = 0.73), while the model without exhibits weaker performance (1% for training, R2 = 0.51).

Inverse Weight-Balancing for Deep Long-Tailed Learning

The performance of deep learning models often degrades rapidly when faced with imbalanced data characterized by a long-tailed distribution. Researchers have found that the fully connected layer trained by cross-entropy loss has large weight-norms for classes with many samples, but not for classes with few samples. How to address the data imbalance problem with both the encoder and the classifier seems an under-researched problem. In this paper, we propose an inverse weight-balancing (IWB) approach to guide model training and alleviate the data imbalance problem in two stages. In the first stage, an encoder and classifier (the fully connected layer) are trained using conventional cross-entropy loss. In the second stage, with a fixed encoder, the classifier is finetuned through an adaptive distribution for IWB in the decision space. Unlike existing inverse image frequency that implements a multiplicative margin adjustment transformation in the classification layer, our approach can be interpreted as an adaptive distribution alignment strategy using not only the class-wise number distribution but also the sample-wise difficulty distribution in both encoder and classifier. Experiments show that our method can greatly improve performance on imbalanced datasets such as CIFAR100-LT with different imbalance factors, ImageNet-LT, and iNaturelists2018.

Abstract 1073: AI-assisted high-dimensional cytological single cell analysis distinguishes normal and malignant urine samples using cell morphology images with high accuracy at the single cell level

Abstract Background: Urothelial cancer (UCC) is the 4th most common malignancy in men. The current clinical problem in UCC management is the high interobserver variability of tissue UCC grading and the low sensitivity of urine cytology for early detection of recurrences and progression. We hypothesize that assessing urine samples and UCC tissues with detailed quantitative image analysis based on artificial intelligence has strong potential in both urine-based early detection of tumor recurrence and progression as well as tissue-based pathological grading. Methods: Here, we utilized the Deepcell® Human Foundation Model (HFM), a self-supervised deep learning model, to assess a cohort of urine samples from 54 normal healthy cases, 19 atypical cases, and 4 UCC cases. High resolution images of label-free single cells were captured and analyzed using the HFM to identify distinct morphological characteristics of cells from different urine origins. The urine samples were loaded onto a single-use microfluidic cartridge as part of the Deepcell® REM-I platform and brightfield images of single cells were captured in real-time. For data analysis, images of all cells run through the REM-I instrument are automatically sent to the Deepcell® Axon Suite, which provides an analysis platform to visualize cell images from specific runs, generate UMAPs of the embedding space, and run additional deep neural network models on samples in silico. Results: High-dimensional analysis of 115 deep learning and morphometric features reveals 12 morphotypes among the pooled urine samples, as represented by morphology UMAP plots. Morphology analysis reveals distinct morphotypes in normal compared to pathological urine samples. Confusion matrix showed deep learning model performance in classifying atypical versus UCC with 89.1% (Atypical (n = 8642), UCC (n = 12637)). This indicates high model performance accuracy for AI prediction of atypical versus UCC samples at the single cell level. Conclusions: The Deepcell® REM-I platform performs well with urine samples and identified numerous cell clusters among normal and pathological urines. AI-assisted high-dimensional morphology analysis distinguishes atypical urine samples from UCC cases with 89.1% accuracy at the single cell level. Future applications of this study include early detection or monitoring of UCC cases using non-invasive patient samples such as urine. Citation Format: Thierry P. van den Bosch, Geert J.L.H. van Leenders, Siamala Sinnadurai, Zhouyang Lian, Ryan Carelli, Matthew Nakaki, Vivian Lu, Manisha Ray, Janifer Cruz, Kelly R. Miller, Mahyar Salek, Maddison Masaeli, Peter J. van der Spek. AI-assisted high-dimensional cytological single cell analysis distinguishes normal and malignant urine samples using cell morphology images with high accuracy at the single cell level [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 1073.

Domain-aware triplet loss in domain generalization

Despite the considerable advances in deep learning for object recognition, there are still several factors that hinder the performance of deep learning models. One of these factors is domain shift, which occurs due to variations in the distribution of the testing and training data. This paper addresses the issue of compact feature clustering in domain generalization, with the aim of optimizing the embedding space from multi-domain data. Specifically, we propose a domain-aware triplet loss for domain generalization, which not only facilitates clustering of similar semantic features but also disperses features that arise from the domain. Unlike previous methods that focus on aligning distributions, our algorithm disperses domain information in the embedding space. Our approach is based on the assumption that embedding features can be clustered based on domain information, which is supported mathematically and empirically in this paper.Furthermore, in our investigation of feature clustering in domain generalization, we observe that the factors that influence the convergence of metric learning loss in domain generalization are more significant than the pre-defined domains. To address this issue, we utilize two methods to normalize the embedding space and reduce the internal covariate shift of the embedding features. Our ablation study illustrates the effectiveness of our algorithm. Additionally, our experiments on benchmark datasets, including PACS, VLCS, and Office-Home, demonstrate that our method outperforms related approaches that focus on domain discrepancy. Notably, our results on RegnetY-16GF are substantially better than state-of-the-art methods on the benchmark datasets. Our code is available at https://github.com/workerbcd/DCT.

A Post-training Framework for Improving the Performance of Deep Learning Models via Model Transformation

Deep learning (DL) techniques have attracted much attention in recent years and have been applied to many application scenarios. To improve the performance of DL models regarding different properties, many approaches have been proposed in the past decades, such as improving the robustness and fairness of DL models to meet the requirements for practical use. Among existing approaches, post-training is an effective method that has been widely adopted in practice due to its high efficiency and good performance. Nevertheless, its performance is still limited due to the incompleteness of training data. Additionally, existing approaches are always specifically designed for certain tasks, such as improving model robustness, which cannot be used for other purposes. In this article, we aim to fill this gap and propose an effective and general post-training framework, which can be adapted to improve the model performance from different aspects. Specifically, it incorporates a novel model transformation technique that transforms a classification model into an isomorphic regression model for fine-tuning, which can effectively overcome the problem of incomplete training data by forcing the model to strengthen the memory of crucial input features and thus improve the model performance eventually. To evaluate the performance of our framework, we have adapted it to two emerging tasks for improving DL models, i.e., robustness and fairness improvement, and conducted extensive studies by comparing it with state-of-the-art approaches. The experimental results demonstrate that our framework is indeed general, as it is effective in both tasks. Specifically, in the task of robustness improvement, our approach Dare has achieved the best results on 61.1% cases (vs. 11.1% cases achieved by baselines). In the task of fairness improvement, our approach FMT can effectively improve the fairness without sacrificing the accuracy of the models.

Effective Forecasting of Key Features in Hospital Emergency Department

Efficiently forecasting the demands within a hospital’s Emergency Department (ED) is critical for optimal resource allocation and patient care management. This study focuses on leveraging deep learning techniques to predict various types of ED patient flows, facilitating informed decision-making by ED managers. The rising success of deep learning networks in modeling timeseries data makes them a compelling choice for patient flow forecasting. In this context, we investigate and compare seven deep learning models-Deep Belief Network (DBN), Restricted Boltzmann Machines (RBM), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), combined GRU and Convolutional Neural Networks (CNN-GRU), LSTM-CNN, and Generative Adversarial Network based on Recurrent Neural Networks (GAN-RNN)—to accurately forecast patient flow within a hospital’s emergency department. To enable traffic flow forecasting, a forecaster layer is introduced for each model. Real-world patient flow data spanning different ED services (biology, radiology, scanner, and echography) at Lille regional hospital in France serve as a case study to evaluate these models. Four effectiveness metrics are employed to assess and compare the forecasting methods. The outcomes demonstrate the superior performance of deep learning models in predicting ED patient flows compared to conventional shallow approaches like ridge regression and support vector regression. Significantly, the Deep Belief Network (DBN) stands out, achieving an averaged mean absolute percentage error of approximately 4.097.

RGB-Angle-Wheel: A new data augmentation method for deep learning models

Deep learning models often rely on a diverse and well-augmented dataset for optimal performance. In this context, the methods of data augmentation are pivotal in boosting the models’ ability to generalize. In this paper, we introduce a novel data augmentation method, which we call RGB-Angle-Wheel, to improve the performance of deep learning models on RGB format images. This method involves rotating each color channel at specific angles to generate new training data that is distinct from the original dataset but shares similar properties. Experimental results on the CIFAR-10, CIFAR-100,and COCO datasets have validated the efficacy of the proposed method for enhancing model performance. Specifically, certain transformations in the red (R) and blue (B) channels improve model accuracy significantly, whereas the effect on the green (G) channel remains limited. These results indicate that the careful selection of transformation parameters plays a critical role in enhancing model performance. The findings of the study indicate that the proposed method can be utilized specifically for image processing, image classification, object detection, and other deep learning applications. Experiments demonstrate that the proposed method improves the model’s efficacy and generalizability.

Deep learning applied to CO2 power plant emissions quantification using simulated satellite images

Abstract. The quantification of emissions of greenhouse gases and air pollutants through the inversion of plumes in satellite images remains a complex problem that current methods can only assess with significant uncertainties. The anticipated launch of the CO2M (Copernicus Anthropogenic Carbon Dioxide Monitoring) satellite constellation in 2026 is expected to provide high-resolution images of CO2 (carbon dioxide) column-averaged mole fractions (XCO2), opening up new possibilities. However, the inversion of future CO2 plumes from CO2M will encounter various obstacles. A challenge is the low CO2 plume signal-to-noise ratio due to the variability in the background and instrumental errors in satellite measurements. Moreover, uncertainties in the transport and dispersion processes further complicate the inversion task. To address these challenges, deep learning techniques, such as neural networks, offer promising solutions for retrieving emissions from plumes in XCO2 images. Deep learning models can be trained to identify emissions from plume dynamics simulated using a transport model. It then becomes possible to extract relevant information from new plumes and predict their emissions. In this paper, we develop a strategy employing convolutional neural networks (CNNs) to estimate the emission fluxes from a plume in a pseudo-XCO2 image. Our dataset used to train and test such methods includes pseudo-images based on simulations of hourly XCO2, NO2 (nitrogen dioxide), and wind fields near various power plants in eastern Germany, tracing plumes from anthropogenic and biogenic sources. CNN models are trained to predict emissions from three power plants that exhibit diverse characteristics. The power plants used to assess the deep learning model's performance are not used to train the model. We find that the CNN model outperforms state-of-the-art plume inversion approaches, achieving highly accurate results with an absolute error about half of that of the cross-sectional flux method and an absolute relative error of ∼ 20 % when only the XCO2 and wind fields are used as inputs. Furthermore, we show that our estimations are only slightly affected by the absence of NO2 fields or a detection mechanism as additional information. Finally, interpretability techniques applied to our models confirm that the CNN automatically learns to identify the XCO2 plume and to assess emissions from the plume concentrations. These promising results suggest a high potential of CNNs in estimating local CO2 emissions from satellite images.