In combination with yearsof experience on controlling the junk MMS, the paper presents a comprehensive study on visual word based similar image retrieval and clustering. Several key points, such as visual word generating, embedding code computing, indexing and scoring are discussed in detail with many excellent engineering practices.%本文总结了不良违规图片管理经验,详细介绍了海量图片相似检索和聚类技术的诸多关键技术,包含视觉词和嵌入码的生成、索引、结果打分等,并给出了许多工程化的实践经验。
展开▼