Focus-Attention Approach in Optimizing DETR for Object Detection from High-Resolution Images

光学(聚焦) 计算机科学 高分辨率 分辨率(逻辑) 对象(语法) 人工智能 计算机视觉 目标检测 遥感 模式识别(心理学) 地理 物理 光学
作者
Nguyen Hoanh,Tran Vu Pham
出处
期刊:Knowledge Based Systems [Elsevier]
卷期号:296: 111939-111939 被引量:3
标识
DOI:10.1016/j.knosys.2024.111939
摘要

Transformer-based detectors have recently achieved remarkable success in object detection, revolutionizing the field with their efficiency and accuracy. However, applying these models to high-resolution images presents significant challenges due to the increased computational demands and complexity of processing dense, high-resolution data. In this paper, we introduce a novel model specifically designed for object detection from high-resolution imagery. This model incorporates a multi-layer object-focus network along with a transformer encoder-decoder structure. Specifically, the model employs a dual-head strategy in the object-focus network to balance the detailed analysis of small objects with computational efficiency. This is achieved by leveraging data sparsity to reduce unnecessary computations in massive background regions. Additionally, to improve detection performance for small objects, we propose a method to effectively apply the transformer encoder-decoder structure combined with the object-focus network on the multi-layer feature maps of the feature pyramid. Our extensive evaluations of the VisDrone, MS-COCO, and UAVid datasets demonstrate that our model outperforms other DETR-based detectors in both detection accuracy and computational speed, highlighting its superior performance. These results indicate a significant advancement in the field of high-resolution object detection utilizing transformer architecture.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
更新
大幅提高文件上传限制,最高150M (2024-4-1)

科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
刚刚
赘婿应助费老三采纳,获得10
刚刚
2秒前
五角星发布了新的文献求助10
2秒前
搜集达人应助Yyy采纳,获得10
2秒前
2秒前
追寻的妙松完成签到,获得积分10
2秒前
3秒前
3秒前
科研通AI2S应助王煊采纳,获得10
4秒前
爆米花应助陈陈采纳,获得10
4秒前
4秒前
4秒前
5秒前
ncjdoi发布了新的文献求助10
6秒前
6秒前
fanny发布了新的文献求助10
7秒前
甜蜜小张完成签到,获得积分10
7秒前
紫苑发布了新的文献求助20
7秒前
叶远望发布了新的文献求助10
8秒前
8秒前
9秒前
沐沐发布了新的文献求助10
9秒前
zhaoxiao发布了新的文献求助10
9秒前
Scarlett发布了新的文献求助10
9秒前
我服有点黑完成签到,获得积分10
10秒前
荣枫发布了新的文献求助10
10秒前
敏感的星星完成签到 ,获得积分10
11秒前
Wt完成签到,获得积分20
11秒前
十一发布了新的文献求助30
12秒前
12秒前
13秒前
zorn应助guoguo采纳,获得20
13秒前
啦啦啦完成签到,获得积分20
14秒前
ncjdoi完成签到,获得积分10
14秒前
15秒前
李健应助木棉采纳,获得10
15秒前
15秒前
15秒前
甜蜜小张发布了新的文献求助10
16秒前
高分求助中
Evolution 10000
ISSN 2159-8274 EISSN 2159-8290 1000
Becoming: An Introduction to Jung's Concept of Individuation 600
Ore genesis in the Zambian Copperbelt with particular reference to the northern sector of the Chambishi basin 500
A new species of Coccus (Homoptera: Coccoidea) from Malawi 500
A new species of Velataspis (Hemiptera Coccoidea Diaspididae) from tea in Assam 500
PraxisRatgeber: Mantiden: Faszinierende Lauerjäger 500
热门求助领域 (近24小时)
化学 医学 生物 材料科学 工程类 有机化学 生物化学 物理 内科学 纳米技术 计算机科学 化学工程 复合材料 基因 遗传学 催化作用 物理化学 免疫学 量子力学 细胞生物学
热门帖子
关注 科研通微信公众号,转发送积分 3160558
求助须知:如何正确求助?哪些是违规求助? 2811730
关于积分的说明 7893251
捐赠科研通 2470605
什么是DOI,文献DOI怎么找? 1315658
科研通“疑难数据库(出版商)”最低求助积分说明 630920
版权声明 602042