System for waste detection and segmentation using combination of vision transformer (ViT) for global feature extraction and mask region-based convolutional neural network (R-CNN) for instance segmentation, in which waste objects are segmented and classified based on learned spatial features
	SIVADARSHAN M R G; SHRIPRASAD S; SARAN V; VARSHAIN G S; MAYURANATHAN M
	2025-04-02
专利权人	SRM VALLIAMMAI ENG COLLEGE (SRMV-Non-standard)
申请日期	2025-04-02
专利号	IN202541032498-A
成果简介	NOVELTY - The system in which waste objects are segmented and classified based on learned spatial features for improved waste identification and sorting is provided. The preprocessing module that applies image resizing, normalization, patch extraction, and data augmentation is utilized to enhance model generalization and segmentation accuracy. The mask R-CNN-based segmentation module generates pixel-wise segmentation masks, thus incorporating Rol Align and bounding box regression to refine object localization and segmentation quality. USE - System for waste detection and segmentation using combination of vision transformer (ViT) for global feature extraction and mask region-based convolutional neural network (R-CNN) for instance segmentation. ADVANTAGE - The vision transformer-based feature extraction mechanism divides images into patches and applies self-attention mechanisms to capture long-range dependencies, thus ensuring improved waste object recognition in cluttered environments. The segmented waste objects are classified using a trained model and evaluated with mean average precision (mAP) to ensure high segmentation accuracy. The system supports real-time waste monitoring, classification, and optimized sorting, helping municipalities, industries, and waste management authorities improve recycling efficiency and reduce environmental impact. The system offers a highly scalable, accurate, and sustainable waste management solution by combining artificial intelligence (AI) powered vision models with robust preprocessing techniques. DESCRIPTION OF DRAWING(S) - The drawing shows a schematic diagram of a system architecture of a system for waste detection and segmentation using combination of vision transformer.
IPC 分类号	G06N-003/045 ; G06N-003/08 ; G06T-007/11 ; G06V-010/774 ; G06V-010/82
国家	印度
专业领域	信息技术
语种	英语
成果类型	专利
文献类型	科技成果
条目标识符	http://119.78.100.226:8889/handle/3KE4DYBR/13219
专题	中国科学院新疆生态与地理研究所
作者单位	SRM VALLIAMMAI ENG COLLEGE (SRMV-Non-standard)
推荐引用方式 GB/T 7714	SIVADARSHAN M R G,SHRIPRASAD S,SARAN V,et al. System for waste detection and segmentation using combination of vision transformer (ViT) for global feature extraction and mask region-based convolutional neural network (R-CNN) for instance segmentation, in which waste objects are segmented and classified based on learned spatial features. IN202541032498-A[P]. 2025.

条目包含的文件
条目无相关文件。

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。