Hybrid Deep Learning Approach for Multi-label Image Classification
کد مقاله : 1117-CFIS (R1)
نویسندگان:
رضا محمدی مقدم *1، حسن ختنلو2، یوسف رضایی3
1گروه آموزشی کامپیوتر- دانشکده مهندسی - دانشگاه بوعلی سینا - همدان - ایران
2گروه آموزشی کامپیوتر - دانشکده مهندسی - دانشگاه بوعلی سینا - همدان - ایران
3گروه آموزشی عمران - دانشکده مهندسی - دانشگاه بوعلی سینا - همدان - ایران
چکیده مقاله:
Multi-label image classification aims to predict multiple labels for a single image which consists of diverse contents. The main challenge in Multi-label classification task to achieve a decent performance is the lack of enough training data. Convolutional Neural Networks (CNN) have shown satisfying results in single-label image classification, but multi-label image classification is still an open field of research. In this paper an efficient hybrid method for multi-label image classification is proposed. The proposed model consists of multiple sub-networks. The experimental results obtained in this study demonstrate the plausible performance of the proposed method on "Pascal VOC 2012" and "Kaggle: Understanding the Amazon from space challenge" datasets.
کلیدواژه ها:
multi-label classification, deep learning, convolutional neural networks, satellite image classification
وضعیت : مقاله برای ارائه شفاهی پذیرفته شده است