Novel Deep Feature Fusion Framework for Multi-Scenario Violence Detection

Computers, Journal Year: 2023, Volume and Issue: 12(9), P. 175 - 175

Published: Sept. 5, 2023

Detecting violence in various scenarios is a difficult task that requires high degree of generalisation. This includes fights different environments such as schools, streets, and football stadiums. However, most current research on detection focuses single scenario, limiting its ability to generalise across multiple scenarios. To tackle this issue, paper offers new multi-scenario framework operates two environments: fighting locations rugby has three main steps. Firstly, it uses transfer learning by employing pre-trained models from the ImageNet dataset: Xception, Inception, InceptionResNet. approach enhances generalisation prevents overfitting, these have already learned valuable features large diverse dataset. Secondly, combines extracted through feature fusion, which improves representation performance. Lastly, concatenation step first scenario with second train machine classifier, enabling classifier both highly flexible, can incorporate without requiring training scratch additional The Fusion model, incorporates fusion models, obtained an accuracy 97.66% RLVS dataset 92.89% Hockey Concatenation model accomplished 97.64% 92.41% datasets just classifier. allows for classification violent within Furthermore, not limited be adapted tasks.

Language: Английский

---

Employing the Concept of Stacking Ensemble Learning to Generate Deep Dream Images Using Multiple CNN Variants

Intelligent Systems with Applications, Journal Year: 2025, Volume and Issue: unknown, P. 200488 - 200488

Published: Jan. 1, 2025

Language: Английский

---

Bridging Psychedelic VR and BCI: Enhancing User Experience through Adaptive EEG-Guided Neural Modulation

Published: Feb. 25, 2025

Language: Английский

---

Neurosymbolic AI and Mechanistic Interpretability: Can They Align in the Artificial General Intelligence Era?

Published: Jan. 1, 2025

Language: Английский

---

A Systematic Review of Deep Dream

Iraqi Journal of Computer Communication Control and System Engineering, Journal Year: 2023, Volume and Issue: unknown, P. 192 - 209

Published: June 29, 2023

Deep Dream (DD) is a new technology that works as creative image-editing approach by employing the representations of CNN to produce dreams-like images taking benefits both and Inception build dream through layer-by-layer implementation. As days go by, DD becomes widely used in artificial intelligence (AI) fields. This paper first systematic review DD. We focused on definition, importance, background, applications Natural language processing (NLP), images, videos, audio are main fields which applied. also discussed concepts DD, like transfer learning Inception. addressed contributions, databases, techniques have been models, limitations, evaluation metrics for each one included research papers. Finally, some interesting recommendations listed serve researchers future. Index Terms— dream, deep CNN, gradient ascent, Inception, style transfer.

Language: Английский

---

Prediction of physical realizations of the coordinated universal time with gated recurrent unit

Review of Scientific Instruments, Journal Year: 2024, Volume and Issue: 95(1)

Published: Jan. 1, 2024

Coordinated Universal Time (UTC), produced by the Bureau International des Poids et Mesures (BIPM), is official worldwide time reference. Given that there no physical signal associated with UTC, realizations of called UTC(k), are very important for demanding applications such as global navigation satellite systems, communication networks, and national defense security, among others. Therefore, prediction differences UTC-UTC(k) to maintain accuracy stability UTC(k) timescales. In this paper, we report first use a deep learning (DL) technique Gated Recurrent Unit (GRU) predict sequence H futures values ten different published on monthly Circular T document BIPM used training samples. We utilize multiple-input, multiple-output strategy. After process where about 300 past difference used, (H = 6) can be predicted using p (typically values. The model has been tested data from When comparing GRU results other standard DL algorithms, found approximation good performance in predicting According our results, error typically 1 ns. frequency instability timescale main limitation reducing prediction.

Language: Английский

---

A Robust Approach for Ulcer Classification/Detection in WCE Images

International Journal of Online and Biomedical Engineering (iJOE), Journal Year: 2024, Volume and Issue: 20(06), P. 86 - 102

Published: April 12, 2024

Wireless Capsule Endoscopy (WCE) is a medical diagnostic technique recognized for its minimally invasive and painless nature the patients. It uses remote imaging techniques to explore various segments of gastrointestinal (GI) tract, particularly hard-to-reach small intestine, making it an effective alternative traditional endoscopic techniques. However, physicians face significant challenge when comes analyzing large number images due effort time required. therefore imperative implement aided-diagnostic systems capable automatically detecting suspicious areas subsequent assessment. In this paper, we present novel approach identify tract abnormalities from WCE images, with particular focus on ulcerated areas. Our involves use Median Robust Extended Local Binary Pattern (MRELBP) descriptor, which effectively overcomes challenges faced image acquisition, such as variations in illumination contrast, rotation, noise. Using machine learning algorithms, conducted experiments extensive Kvasir-Capsule dataset, subsequently compared our results recent relevant studies. Noteworthy fact that achieved accuracy 97.04% SVM (RBF) classifier 96.77% RF classifier.

Language: Английский

---

Deep network fault diagnosis for imbalanced small-sized samples via a coupled adversarial autoencoder based on the Bayesian method

Review of Scientific Instruments, Journal Year: 2024, Volume and Issue: 95(5)

Published: May 1, 2024

Deep network fault diagnosis methods heavily rely on abundant labeled data for effective model training. However, small-sized samples and imbalanced often lead to insufficient features, resulting in accuracy degradation even instability the model. To address this challenge, paper introduces a coupled adversarial autoencoder (CoAAE) based Bayesian method. This aims solve issue of by generating fake integrating them with original ones. Within CoAAE framework, probability density distribution is captured using an encoder are generated random sampling from decoding them. process interaction between classifier obtain prior encoder’s parameters. The parameters updated through decoder’s reconstruction process, leading posterior distribution. Concurrently, decoder trained enhance its ability reconstruct accurately. imbalance samples, parallel employed. shares weights extraction layer encoder, enabling it learn joint fault-related normal samples. evaluate effectiveness proposed augmentation method, experiments were conducted bearing database Case Western Reserve University ResNet18 as deep learning representative. results demonstrate that can effectively augment datasets outperform other advanced methods.

Language: Английский

---

Amplifying the Anomaly: How Humans Choose Unproven Options and Large Language Models Avoid Them

Creativity Research Journal, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 22

Published: June 17, 2024

Both large language models (LLMs) and the human brain develop internal of reality to make accurate predictions. typically prefer choices with strongest track records. However, when faced a creative challenge, LLMs remain committed high-probability options while humans can opt for unproven ones. This paper delves into one way making unlikely events plausible—"amplifying anomaly." The concept involves extrapolating viable consequences from an proposition. Rather than being treated as oddball or "one-offs," anomaly permeates work. Notably, novelty appropriateness be in tension each other, high utility coming at cost low novelty. Amplifying aligns these competing demands. It enhances originality: rarer proposition more thoroughly it is worked out, unique surprising result. At same time, effectiveness value option also rises: thorough elaboration product establishes its fitness. Musical examples by Beethoven, Schubert, contemporary composer Sky Macklay, along products other domains, illustrate this principle. Classic have several limitations that difficult amplify anomaly: they are steered toward norm-driven outcomes, short-term decisions, not designed self-evaluate. As result, difficulty developing unusual propositions non-obvious without guidance. Alternatives approaches, including adversarial networks team AI, briefly examined. Implications future computational creativity discussed.

Language: Английский

---

Deep Learning-Based Fire Detection for Enhanced Safety Systems

Wasit Journal of Pure sciences, Journal Year: 2023, Volume and Issue: 2(4), P. 45 - 55

Published: Dec. 30, 2023

Fire detection systems are a critical aspect of modern safety and security systems, playing pivotal role in safeguarding lives property against the destructive force fires. Rapid accurate identification fire incidents is essential for timely response mitigation efforts. Traditional methods have made substantial advancements, but with advent computer vision technologies, field has witnessed transformative shift. This paper presents method using deep convolutional neural network (CNN) models. approach used transfer learning by employing two pre-trained CNN models from ImageNet dataset: VGG (Visual Geometry Group) InceptionV3 to extract valuable features input images. Then, these extracted serve as machine (ML) classifier, namely Softmax classifier. The activation function computes probability distribution assign class probabilities discriminating between types images: non-fire. Experimental results showed that proposed successfully detected areas achieved seamless classification performance compared other current methods.

Language: Английский

---