1 views collected around this technical thread.
The article explains how an e‑commerce risk‑control team leverages OpenAI's CLIP model to generate image and text embeddings and stores them in the Milvus cloud‑native vector database to enable fast, scalable similarity searches for compliance verification and risk detection.
This article reviews the fundamentals, feature extraction methods, evaluation metrics, and common datasets of content‑based image retrieval (CBIR), discussing traditional low‑level features, local descriptors, unsupervised and supervised learning approaches, and recent deep‑learning models for improving retrieval performance.
This article shares the development of a global search platform that leverages AI technologies such as Chinese word segmentation, part‑of‑speech tagging, text similarity via Simhash and Synonyms, image similarity using histogram, Hamming distance and ResNet‑50, and multimodal CLIP‑based models to improve search efficiency and accuracy.
This article presents a complete image‑search solution that extracts visual features with the VGG16 model, stores them in the Milvus vector database, and provides a set of web APIs for training, querying, counting, searching, and deleting image vectors, all deployed via Docker containers.
This article presents a comprehensive study of multi‑label image recognition applied to 58.com’s business scenarios, covering problem motivation, dataset construction, evaluation metrics, mainstream deep‑learning methods, an asymmetric‑loss‑based optimization pipeline, and practical output schemes for recommendation and retrieval.
To efficiently filter unchanged POI signboards, the authors propose a multimodal image‑retrieval system that combines enhanced global and local visual features with BERT‑encoded OCR text, using metric learning and alignment techniques to achieve over 95 % accuracy while handling occlusion, viewpoint variation, and subtle text changes.
The article summarizes five Ant Financial papers accepted at WWW2020, covering an efficient sentence‑similarity model (Enhanced‑RCNN), a graph‑based multi‑turn dialogue system (IntentDial), a low‑label captcha recognizer, a solver for billion‑scale knapsack problems, and a novel equal‑distance/equal‑distribution triplet loss for image retrieval.
This article summarizes a team's practical experience with various image feature extraction methods—including global, local, and CNN features—and demonstrates their use in image retrieval and no‑reference quality assessment through extensive experiments and analysis.