A novel image retrieval strategy based on transfer learning and hand-crafted features for wool fabric
作者：
Highlights：
• Propose a novel retrieval strategy by transfer learning and hand-crafted features.
• Shrink the search space by the pre-trained model with label smoothing.
• Characterize delicate features by hand-crafted features in shrunk retrieval pools.
• Build the database which contains 82,073 woolen fabric images for experiments.
• Attach the process sheet to the image for industrial application.
摘要
•Propose a novel retrieval strategy by transfer learning and hand-crafted features.•Shrink the search space by the pre-trained model with label smoothing.•Characterize delicate features by hand-crafted features in shrunk retrieval pools.•Build the database which contains 82,073 woolen fabric images for experiments.•Attach the process sheet to the image for industrial application.
论文关键词：Pre-trained model,Label smoothing,Retrieval strategy,Delicate features
论文评审过程：Received 16 January 2021, Revised 11 October 2021, Accepted 12 November 2021, Available online 25 November 2021, Version of Record 7 December 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.116229
