160 likes | 330 Views
Face Detection And Recognition For Distributed Systems. Meng Lin and Ermin Hod žić. Motivation. Security systems Digital cameras, adjustments Social networks Marketing. Fac e Detection. Locating and extracting faces in images Rectangles as output. Face Recognition.
E N D
Face Detection And Recognition ForDistributed Systems Meng Lin and ErminHodžić
Motivation • Security systems • Digital cameras, adjustments • Social networks • Marketing
Face Detection • Locating and extracting faces in images • Rectangles as output
Face Recognition • Identifying person on image • Finding closest match
MapReduce model • Parallel model • Framework for distributed systems • Hadoop local filesystem • Amazon Elastic MapReduce
Solution: Face Detection • Partitions based on face scale
Solution: Face Recognition • Distribution of recognizers • Recognizing in parallel • Reduce the most confident result
Hadoop Image Handling • Images as text • Collection of images in a big file • Utilize Hadoop default input format
Evaluation • Segmented Face Detection • Raw OpenCV Face Detection
Evaluation • Face recognition • More nodes = slower • Input and data transfer overhead • Jobs computationally cheap
Conclusion • Easily scalable parallel mode • Generalized framework • OpenCV just one sample tool • A lot of communication • Low utilization of processors