The application of improved densenet algorithm in accurate image recognition Scientific Reports

Thermal fault diagnosis of complex electrical equipment based on infrared image recognition Scientific Reports

ai based image recognition

No picture enhancer software was employed for additional processing of the captured images, ensuring a true-to-life acquisition to the greatest extent possible. Confidence intervals and standard deviations for AUROC were computed via the Delong method60. All other confidence intervals, standard deviations, and p-values were computed via bootstrapping with 2000 samples.

Drone image recognition and intelligent power distribution network equipment fault detection based on the transformer model and transfer learning – Frontiers

Drone image recognition and intelligent power distribution network equipment fault detection based on the transformer model and transfer learning.

Posted: Thu, 29 Aug 2024 07:00:00 GMT [source]

These results suggest that the identified subgroup based on histopathology images is biologically distinct. Furthermore, our gene expression analysis revealed the upregulation of PI3k-Akt, Wnt, and Cadherin signaling pathways both in p53abn-like NSMP and p53abn groups (compared to NSMP). All these results suggest genomic and transcriptomic similarities between the p53abn-like NSMP and p53abn cases and potential defects in the DNA damage repair process as a possible biological mechanism. This observation suggests that they may have different etiologies and warrants further biological interrogation of these groups in future studies.

Data availability statement

Brain tumors are a major public health problem in the healthcare sector, and accurate diagnosis, treatment, and follow-up processes are critical. AI has become an important tool for improving these processes and has great potential for early diagnosis and treatment of brain tumors. Moreover, AI technology assists in decluttering digital libraries by identifying duplicates and unnecessary images, which simplifies media management and frees up storage space.

  • After all the required images have been captured, they are sent to the image preprocessing stage to be adjusted before use.
  • For further data augmentation, a slightly blurred vision of the grayscale image was created, and the aforementioned thresholding techniques were also applied.
  • Search results may include related images, sites that contain the image, as well as sizes of the image you searched for.
  • These diseases are Black scurf, common scab, black leg, pink rot etc. are caused by different causative agents.
  • D In order to predict slide-level labels, the extracted features are fed into the VLAD aggregation method.

Inception v3 is the third version of the series with additional factorization convolutions, aiming to reduce the number of parameters while maintaining network efficiency. In addition to this, several other techniques for optimizing the network have been suggested to loosen the constraints for more straightforward model adaptation. These techniques include regularization, dimension reduction, and parallelized computations.

In the information age, IR technology has demonstrated powerful practical functions, and how to establish better IR and classification models has attracted more and more experts’ attention. To promote the accuracy of distinguishing cashmere and wool, Zhu et al. established an IR model based on multi feature selection and random forest. The model utilized a combination of correlation, principal component analysis, and weight coefficients to select important and sensitive features.

Revolutionizing agriculture with artificial intelligence: plant disease detection methods, applications, and their limitations

According to the findings, CNorm outperformed Base, while AIDA is the most effective method in separating the five histotypes of ovarian cancer. Similarly, the representation of the feature space of the Pleural and Bladder datasets showed that AIDA outperformed the other approaches in generating more discriminative features for subtype classification. Overall, across all datasets, Macenko, CNorm, and ADA consistently improved the performance of the target datasets to a greater extent than HED, with a notable margin. Furthermore, all methods demonstrated minimal impact on the performance of the source domain. Visual microscopic study of diseased tissue by pathologists has been the cornerstone in cancer diagnosis and prognostication for more than a century. Training loss measures the error on the training data, while validation loss evaluates the model’s performance error on an independent dataset.

Educators can devise targeted teaching improvement strategies by identifying key verbal communication indicators, such as adjusting speech speed or enhancing speech comprehensibility, to elevate students’ learning experiences. Personalized learning experiences, especially in aspects like speech speed and content similarity, will aid students in better assimilating into the online learning environment, aligning more closely with subject interests and learning styles. Ultimately, this contributes to refining individual educators’ teaching methods and provides valuable insights for the entire education system’s development.

The second approach does not involve changes to model training, but instead to score threshold selection at model inference. As the AI model outputs a continuous score from 0 to 1 for the “No Findings” vs. “Findings Present” task, a threshold must be chosen to generate binary outputs. This threshold is typically chosen based on a target metric and the model’s predictions across a validation set. Given the view position results above, we asked if separate thresholds for each view could help mitigate the underdiagnosis bias. The thresholds would again be calculated in the validation set, but separately for each view instead of having one single threshold across all views.

This visualization is also available for representative malignant cases within the Pleural and Bladder cancer datasets (Figs. 10 and 11). In the pleural cancer cases, the top three patches showed high cellularity with densely packed spindle cells, while the low-ranked patches were much more paucicellular and featured areas of collagen. In bladder cancer, the top three patches selected by the method contained subtype-specific histologic features including tumor epithelium, while the bottom three patches primarily encompassed nonspecific stromal or necrotic areas. For example, the most discriminative areas within the top three patches demonstrate the presence of multiple tumor cell clusters within the same lacuna with peripherally oriented nuclei, a typical feature of micropapillary urothelial carcinoma58. Heatmap analysis of samples (a) from the source domain and (b, c) from the target domain of the Pleural cancer dataset. The first column is the input slide incorporating the tumor annotation provided by the pathologist, and the second to fourth columns are the outputs of Base, CNorm, and AIDA methods.

Zooming in on the images allows the model to noticeably identify the loop structure of the images. Our results show that these cases (referred to as p53abn-like NSMP) have inferior outcomes compared to the other NSMP ECs, similar to that of p53abn EC, in three independent cohorts. Furthermore, shallow whole genome sequencing studies suggested that the genomic architecture of the p53abn-like NSMP differs from other NSMP ECs, showing increased copy number abnormalities, a characteristic of p53abn EC. Endometrial cancer (EC) has four molecular subtypes with strong prognostic value and therapeutic implications.

Conversely, ADA with the CTransPath backbone exhibited superior performance when trained with augmentation. The distinguishing factor between AIDA and ADA lies in the inclusion of the FFT-Enhancer module. Our findings indicate that when utilizing a backbone with domain-specific pre-trained weights, the FFT-Enhancer can enhance model performance without augmentation, surpassing its augmented counterpart. This outcome may be attributed to CTransPath’s extensive training on a diverse array of histopathology images, enabling adaptation to various general variations, including those related to color. Consequently, the pre-trained weights enable the model to accommodate samples with distinct color spaces, with the FFT-Enhancer aiding in sharpening the focus on tumor morphology and shape during training.

ai based image recognition

We hypothesized that such cases may in fact exhibit similar clinical behavior as p53abn ECs. Following Seyyed-Kalantari et al., we evaluate the diagnostic AI models on the binary task of classifying if findings ChatGPT App are present using the “No Findings” label available in each dataset1. As the AI model outputs a continuous 0–1 score, a threshold must be chosen to binarize the model’s outputs, which is described further below.

Examples include regular cleaning of camera lenses to prevent biofouling, installation on farms in a way that is feasible and at a reasonable cost, and robust power supplies and data communication devices. Investigating the decline in scallop production requires accurate data and information, but underwater observations can be time-consuming, challenging and often unreliable. With this in mind, researchers at the Hokkaido Research Organization have come up with a unique AI image-recognition technique to monitor scallops and study conditions underwater to identify potential causes of abnormal growth and mortality. That’s remarkable in itself, but compare this to our experience with training CNNs in December 2015, 36 months back when automated image detection in the ImageNet competition first exceeded human performance. Roughly another 18 months before this chart begins Microsoft researchers used an extraordinary 152 layer CNN which was five times the size of any previous system to beat the human benchmark for the first time (97% for the CNN, 95% for humans).

Comparison of NSMP and p53abn-like NSMP

Recently, self-supervised auxiliary tasks have been utilized to improve the performance of these networks in the context of histopathology images13,32,41,42. This case application demonstrates the feasibility and accuracy of integrating rock type identification, weathering degree assessment, and correction factor application in practical engineering. The method not only enhances the precision of rock strength prediction but also provides a reliable scientific basis for tunnel construction design and support structure selection, thereby improving the safety and economy of the project. Additionally, this case highlights the advantages of combining modern neural network technology with traditional geotechnical engineering knowledge, showcasing the importance of technological innovation in engineering practice.

Secondly, deep learning models have been shown to recognize the tissue submission site even after deploying color normalization techniques. This was shown in a study by Howard et al.34, where they analyzed the differences in slide image characteristics from different centers using classical descriptive statistics. Their study revealed that all these statistics exhibited variance according to the tissue submitting center while color normalization methods could improve only some of these statistical characteristics and had no effect on the remainder. This suggests that these techniques do not necessarily remove all the site-specific signatures and therefore, may not lead to more generalizable models. Deep learning models tend to be data-intensive and require a significant amount of training data. In an ideal scenario, a network should be trained using data acquired from a single center, and subsequently applied to multiple centers.

WGF is employed to process the input image, yielding a smoother base layer, and the detail layer image is obtained by subtracting this base layer from the original image, as illustrated in the following equations17. In recent times, visual search has revolutionised the way we shop online, making the hunt for products as smooth as a picture. This innovative tool allows customers to search for items using images instead of text, providing an additional intuitive shopping experience. Let’s look into the profound impact and benefits of AI-powered image recognition and visual search on ecommerce. 5, the classification performance is high for all four models (VGG16 and VGG19 models, CNN model, EfficientNetB4 model, InceptionV3 model).

Using the polygon annotation method in the LabelMe annotation software, we perform pixel-level annotation of all rock properties and backgrounds in the images, as shown in Fig. The upper part shows the image data with red boundary lines indicating the annotated areas; the lower part shows the corresponding images after polygon annotation. In summary, exploring and developing AI and neural network-based methods for rock strength assessment has become a key direction for addressing this issue.

The Performance assessment of single-stage Object detection algorithms as shown in Figure 3. In the task of object detection, a dataset with strong applicability can effectively test and assess the performance of the algorithm and promote the development of research in related fields. Scholars have extensively researched educational data mining, online courses, online course teaching quality, educators’ teaching characteristics, and TBA, both theoretically and practically. There is a research gap in secondary school education-oriented classroom discourse analysis (CDA). Notably, the particularity of teaching methods in the secondary school teaching environment has been considered in sporadic cases. However, their research focuses on the expressive skills and techniques of classroom discourse, providing a reference for this work.

Although many patients with endometrial carcinoma may be cured by surgery alone, about 1 in 5 patients have more aggressive disease and/or have the disease spread beyond the uterus at the time of diagnosis. Identifying these at-risk individuals remains a challenge, with current tools lacking precision. Molecular classification offers an objective and reproducible classification system that has strong prognostic value; improving the ability to discriminate outcomes compared to conventional pathology-based risk stratification criteria. However, it has become apparent that within molecular subtypes and most profoundly within NSMP ECs, there are clinical outcome outliers. The current study addresses this diversity by employing AI-powered histopathology image analysis, in an attempt to identify clinical outcome outliers within the most common molecular subtype of endometrial cancer (Fig. 5). With respect to reducing the underdiagnosis bias, the results are again consistent as the view-specific threshold approach reduces this bias in MXR across all strategies (Supplementary Fig. 3).

Specifically, AIDA achieved balanced accuracy scores of 80.93%, 72.95%, 63.42%, and 75.23% for the Ovarian, Pleural, Bladder, and Breast datasets, respectively. This demonstrates AIDA’s superior robustness and effectiveness compared to ADA in enhancing feature extraction capabilities, irrespective of the network’s initial weights. You can foun additiona information about ai customer service and artificial intelligence and NLP. In order to assess the efficacy and utility of different layers as feature extractors, we constructed a domain classifier exploiting the output of the Xth convolutional block. In the initial three datasets, AIDA-4 exhibited superior performance in target-domain classification, except for the Breast dataset, where AIDA-5 outperformed it. However, the performance gap for the Breast dataset was minimal, with an estimated difference of approximately 1%, indicating that both AIDA-4 and AIDA-5 exhibited comparable performance on this dataset. This suggests that the fourth convolutional block contributes to more generalizable and optimal features for the domain classifier.

A thermal fault diagnosis method for electrical equipment based on the DeeplabV3 + semantic segmentation model is introduced, which leverages temperature differences for fault determination. This study proposes a comprehensive method ranging from preprocessing to recognition to thermal fault diagnosis of infrared images, offering practical insights and robust solutions for automating the ChatGPT infrared inspection of electrical equipment. The daily inspection of power equipment generates a massive amount of infrared images. It remains necessary to manually assess whether the equipment exhibits temperature abnormalities10. This method, only suitable for analyzing and diagnosing a limited number of image tasks, cannot cope with the detection of a large volume of infrared images.

ai based image recognition

This approach explores a scientifically sound method for calculating and analyzing the four indicators of online classroom discourse. The structured calculation of these four indicators is realized, as depicted in Fig. The online course-oriented data mining technology based on AI targets the unique data collected from the teaching environment, teaching objects, and teaching process in online courses. It focuses on big data in online courses, which falls into the main category of educational big data research and application8. Recently, the application and research of educational data mining technology in online courses have been increasing.

In contrast, the disease name, diseased image, and unique symptoms that damage specific tomoato plant parts are highlighted (Table 4). Furthermore, the detailed explanations of the previous studies to predict the tomato diseases automatically are provided below. ● We discussed the AI based plant disease classification, where, the automated approaches to classify disease in each respective vegetable are provided. A data classification ai based image recognition policy is vital in AI data classification as it outlines the criteria to categorize and manage various types of data within your organization. It plays a vital role in ensuring appropriate protection measures are in place, which becomes especially critical when training AI models with sensitive data. Get a free data classification policy template and learn how to create your own by reading our data classification policy article.

As Natural Language Processing (NLP) technology has progressed, additional methods such as stem extraction, stop word removal, and part-of-speech tagging have been integrated into Text Similarity Measurement (TSM). Contemporary TSM methods often combine semantic information with various weighting, regularization strategies, and NLP techniques29. Video is a multimedia resource combining visual and auditory elements, with the teaching video carrying the main instructional content of the course. Therefore, recognizing speech in the teaching video allows for the extraction of semi-structured classroom discourse text. On the other hand, teaching courseware (teaching content) is predominantly conveyed through the visual channel.

The related literature of the GoogLeNet network is a typical optimization method of the Inception module (Shi et al., 2017) and the optimization process is shown in Figure 6. The generative adversarial network or reinforcement learning-related technologies required for unsupervised data augmentation methods are complex and diverse, which hinders researchers’ exploration. Table 4 Data Augmentation-based object detection in Multimedia, Agriculture and Remote sensing. Table 2 Advantages, disadvantages, and applicable scenarios of two-stage Object detection algorithms. We also tested for external validity of models across different unseen datasets, as well as with different datasets in combination. Looking ahead, the researchers are not only focused on exploring ways to enhance AI’s predictive capabilities regarding image difficulty.

The ML models then predict whether a leaf is healthy or diseased (Ayaz et al., 2019). The framework with predetermined steps to predict the plant disease is presented (Figure 1). This paper realizes infrared image denoising, recognition, and semantic segmentation for complex electrical equipment and proposes a thermal fault diagnosis method that incorporates temperature differences.

Leave a Comment