Simplified Knowledge Distillation for Deep Neural Networks Bridging the Performance Gap with a Novel Teacher–Student Architecture

Electronics, Journal Year: 2024, Volume and Issue: 13(22), P. 4530 - 4530

Published: Nov. 18, 2024

The rapid evolution of deep learning has led to significant achievements in computer vision, primarily driven by complex convolutional neural networks (CNNs). However, the increasing depth and parameter count these often result overfitting elevated computational demands. Knowledge distillation (KD) emerged as a promising technique address issues transferring knowledge from large, well-trained teacher model more compact student model. This paper introduces novel method that simplifies process narrows performance gap between models without relying on intricate representations. Our approach leverages unique network architecture designed enhance efficiency effectiveness transfer. Additionally, we introduce streamlined transfers effectively through simplified process, enabling achieve high accuracy with reduced Comprehensive experiments conducted CIFAR-10 dataset demonstrate our proposed achieves superior compared traditional KD methods established architectures such ResNet VGG networks. not only maintains but also significantly reduces training validation losses. Key findings highlight optimal hyperparameter settings (temperature T = 15.0 smoothing factor α 0.7), which yield highest lowest loss values. research contributes theoretical practical advancements distillation, providing robust framework for future applications compression optimization. simplicity pave way accessible scalable solutions deployment.

Language: Английский

---

NeuroInsight: a revolutionary self-adaptive framework for precise brain tumor classification in medical imaging using adaptive deep learning

Signal Image and Video Processing, Journal Year: 2025, Volume and Issue: 19(2)

Published: Jan. 3, 2025

Language: Английский

---

Low-cost self-constructing multi-objective multi-mode parallel vestibular schwannoma recognition method

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 108, P. 107964 - 107964

Published: April 19, 2025

Language: Английский

---

Deep Learning Approaches for Brain Tumor Detection and Classification Using MRI Images (2020 to 2024): A Systematic Review

Deleted Journal, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 30, 2024

Brain tumor is a type of disease caused by uncontrolled cell proliferation in the brain leading to serious health issues such as memory loss and motor impairment. Therefore, early diagnosis tumors plays crucial role extend survival patients. However, given busy nature work radiologists aiming reduce likelihood false diagnoses, advancing technologies including computer-aided artificial intelligence have shown an important assisting radiologists. In recent years, number deep learning-based methods been applied for detection classification using MRI images achieved promising results. The main objective this paper present detailed review previous researches field. addition, This summarizes existing limitations significant highlights. study systematically reviews 60 articles published between 2020 January 2024, extensively covering transfer learning, autoencoders, transformers, attention mechanisms. key findings formulated provide analytic comparison future directions. aims comprehensive understanding automatic techniques that may be useful professionals academic communities working on detection.

Language: Английский

---

Deep Learning-Based Classification of Macrofungi: Comparative Analysis of Advanced Models for Accurate Fungi Identification

Sensors, Journal Year: 2024, Volume and Issue: 24(22), P. 7189 - 7189

Published: Nov. 9, 2024

This study focuses on the classification of six different macrofungi species using advanced deep learning techniques. Fungi species, such as Amanita pantherina, Boletus edulis, Cantharellus cibarius, Lactarius deliciosus, Pleurotus ostreatus and Tricholoma terreum were chosen based their ecological importance distinct morphological characteristics. The research employed 5 machine techniques 12 models, including DenseNet121, MobileNetV2, ConvNeXt, EfficientNet, swin transformers, to evaluate performance in identifying fungi from images. DenseNet121 model demonstrated highest accuracy (92%) AUC score (95%), making it most effective distinguishing between species. also revealed that transformer-based particularly transformer, less effective, suggesting room for improvement application this task. Further advancements could be achieved by expanding datasets, incorporating additional data types biochemical, electron microscopy, RNA/DNA sequences, ensemble methods enhance performance. findings contribute valuable insights into both use biodiversity conservation

Language: Английский

---

An Advanced Brain Tumor Detection Model Using a Hybrid (1D/2D) Convolution-Based Efficient Attention Network with Image Feature Extraction

Sensing and Imaging, Journal Year: 2025, Volume and Issue: 26(1)

Published: April 16, 2025

Language: Английский

---

NeuroInsight: A Revolutionary Self-Adaptive Framework for Precise Brain Tumor Classification in Medical ImagingUsing Adaptive Deep Learning

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: March 13, 2024

This study presents a robust framework for the classification of brain tumors, beginning with meticulous data curation from 233 patients. The dataset comprises diverse range T1-weighted contrast-enhanced images, encompassing meningioma, glioma, and pituitary tumor types. Rigorous organization, pre-processing, augmentation techniques are applied to optimize model training. proposed self-adaptive incorporates cutting-edge algorithm, leveraging Adaptive Contrast Limited Histogram Equalization (CLAHE) Self-Adaptive Spatial Attention. CLAHE enhances grayscale images by tailoring contrast unique characteristics each region. Attention, implemented through an Attention Layer, dynamically assigns weights spatial locations, thereby enhancing sensitivity critical regions. architecture integrates transfer learning models, including DenseNet169, DenseNet201, ResNet152, InceptionResNetV2, contributing its robustness. DenseNet169 serves as feature extractor, capturing hierarchical features pre-trained weights. Model adaptability is further enriched components such batch normalization, dropout, layer adaptive rate strategy, mitigating overfitting adjusting rates during Technical details, use Adam optimizer softmax activation function, underscore model's optimization multi-class capabilities. model, which amalgamates mechanisms, emerges powerful tool detection in medical imaging. Its nuanced comprehension facilitated attention positions it promising advancement computer-aided diagnosis neuroimaging. Leveraging DenseNet201 mechanism, surpasses previous methods, achieving accuracy 94.85%, precision 95.16%, recall 94.60%, showcasing potential enhanced generalization challenging realm image analysis.

Language: Английский

---

Brain Tumor Detection with YOLOv8

Published: June 21, 2024

Language: Английский

---