18054524. Clustering Images for Anomaly Detection simplified abstract (Google LLC)
Contents
Clustering Images for Anomaly Detection
Organization Name
Inventor(s)
Kihyuk Sohn of Mountain View CA (US)
Jinsung Yoon of San Jose CA (US)
Chun-Liang Li of Mountain View CA (US)
Tomas Jon Pfister of Foster City CA (US)
Chen-Yu Lee of Mountain View CA (US)
Clustering Images for Anomaly Detection - A simplified explanation of the abstract
This abstract first appeared for US patent application 18054524 titled 'Clustering Images for Anomaly Detection
Simplified Explanation
The abstract describes a computer-implemented method for clustering images into groups based on their patch embeddings. The method involves receiving a request to assign images into groups, obtaining a set of images, extracting patch embeddings from each image, calculating distances between the patch embeddings of each image, and assigning the images into groups based on these distances.
- The method receives a request to cluster images into groups.
- A set of images is obtained for processing.
- Patch embeddings are extracted from each image.
- Distances between the patch embeddings of each image are calculated.
- Images are assigned into groups based on these distances.
Potential Applications
- Image classification and organization
- Anomaly detection in image datasets
- Content-based image retrieval systems
Problems Solved
- Efficient clustering of images into groups
- Identification of anomalies or outliers within image datasets
- Simplifying image organization and retrieval processes
Benefits
- Improved efficiency in image clustering and organization
- Enhanced anomaly detection capabilities in image datasets
- Streamlined content-based image retrieval systems
Original Abstract Submitted
A computer-implemented method includes receiving an anomaly clustering request that requests data processing hardware to assign each image of a plurality of images into one of a plurality of groups. The method also includes obtaining a plurality of images. For each respective image, the method includes extracting a respective set of patch embeddings from the respective image, determining a distance between the respective set of patch embeddings and each other set of patch embeddings, and assigning the respective image into one of the plurality of groups using the distances between the respective set of patch embeddings and each other set of patch embeddings.
