Published: Aug. 8, 2024
Language: Английский
Computer Science Review, Journal Year: 2025, Volume and Issue: 56, P. 100723 - 100723
Published: Jan. 9, 2025
Language: Английский
Citations
1
Sustainability, Journal Year: 2025, Volume and Issue: 17(8), P. 3502 - 3502
Published: April 14, 2025
Municipal solid waste (MSW) accumulation is a critical global challenge for society and governments, impacting environmental social sustainability. Efficient separation of MSW essential resource recovery advancing sustainable urban management practices. However, manual classification remains slow inefficient practice. In response, advances in artificial intelligence, particularly machine learning, offer more precise efficient alternative solutions to optimize this process. This research presents the development light deep neural network called R3sNet (three “Rs” Reduce, Reuse, Recycle) with residual modules trained end-to-end binary MSW, capability faster inference. The results indicate that combination processing techniques, optimized architecture, training strategies contributes an accuracy 87% organic 94% inorganic waste. outperforms pre-trained ResNet50 model by up 6% both while also reducing number hyperparameters 98.60% GFLOPS 65.17% compared ResNet50. sustainability improving processes, facilitating higher recycling rates, landfill dependency, promoting circular economy. model’s computational requirements translate into lower energy consumption during inference, making it well-suited deployment resource-constrained devices smart environments. These advancements support following Sustainable Development Goals (SDGs): SDG 11: Cities Communities, 12: Responsible Consumption Production, 13: Climate Action.
Language: Английский
Citations
0
Algorithms for intelligent systems, Journal Year: 2025, Volume and Issue: unknown, P. 107 - 119
Published: Jan. 1, 2025
Language: Английский
Citations
0
Frontiers in Computer Science, Journal Year: 2025, Volume and Issue: 7
Published: April 9, 2025
Significant advancements in object detection have transformed our understanding of everyday applications. These developments been successfully deployed real-world scenarios, such as vision surveillance systems and autonomous vehicles. Object recognition technologies evolved from traditional methods to sophisticated, modern approaches. Contemporary systems, leveraging high accuracy promising results, can identify objects interest images videos. The ability Convolutional Neural Networks (CNNs) emulate human-like has garnered considerable attention. This study provides a comprehensive review evaluation CNN-based techniques, emphasizing the deep learning that significantly improved model performance. It analyzes 1-stage, 2-stage, hybrid approaches for recognition, localization, classification, identification, focusing on CNN architecture, backbone design, loss functions. findings reveal while 2-stage achieve superior precision, 1-stage offer faster processing lower computational complexity, making them advantageous specific real-time
Language: Английский
Citations
0
Energies, Journal Year: 2024, Volume and Issue: 17(20), P. 5177 - 5177
Published: Oct. 17, 2024
Within the area of environmental perception, automatic navigation, object detection, and computer vision are crucial demanding fields with many applications in modern industries, such as multi-target long-term visual tracking automated production, defect driverless robotic vehicles. The performance has greatly improved recently thanks to developments deep learning algorithms hardware computing capabilities, which have spawned creation a large number related applications. At same time, rapid increase autonomous systems market, energy consumption become an increasingly critical issue SLAM (Simultaneous Localization Mapping) algorithms. This paper presents results detailed review over 100 papers published course two decades (1999–2024), primary focus on technical advancement vision. To elucidate foundational principles, examination typical based traditional correlation filtering was initially conducted. Subsequently, comprehensive overview state-of-the-art advancements learning-based techniques compiled. Furthermore, comparative analysis conventional novel undertaken discuss future trends directions Lastly, feasibility employing context vehicles explored. Additionally, intelligent robots for low-carbon, unmanned factories, we discussed model optimization pruning quantization, highlighting their importance enhancing efficiency. We conducted comparison various algorithms, exploration how balance these factors discussion potential development trends.
Language: Английский
Citations
3
International Journal of Integrated Science and Technology, Journal Year: 2024, Volume and Issue: 2(5), P. 494 - 505
Published: June 1, 2024
In an era of increasing concern for environmental sustainability, waste management remains important global issue. Efficient classification, in particular distinguishing between organic and recyclable materials, is essential reducing impact. Traditional manual classification methods are often error-prone inefficient. This research evaluates the performance SVM models with RBF Polynomial kernels using SqueezeNet feature extraction. Datasets from Kaggle were preprocessed augmented to improve model training. The experimental results show that kernel outperforms classifying waste, accuracy 97.9% compared 97.3% kernel. finding underscores importance selection parameter tuning optimising non-linear tasks. contributes development more efficient accurate technologies, promoting better practices. Further recommended explore advanced extraction expand scope cover a wider range categories.
Language: Английский
Citations
0
Published: Aug. 8, 2024
Language: Английский
Citations
0
Published: Aug. 8, 2024
Language: Английский
Citations
0