VisageCloud merges state-of-the-art deep learning algorithms for face recognition and classification with data querying, tagging and querying techniques so as to empower you to leverage the full value of your data.
Face recognition meets big data. In cloud or on-premise.
2. Agenda
Context
Use Cases
Demo: what actors you look like?
Features
Face Detection
Classification
Feature Extraction & Analysis
Competition Analysis
Roadmap
Technical Performance
Offering
3. The Context
Market size estimated* to reach 6.8 – 9.6 billion USD worldwide by 2022
Machine face recognition comparable to human capabilities
Computational effort required for analysis economically feasible
Face analysis has attracted big players from the industry (Amazon, Microsoft)
4. Use Cases
Online dating: finding members who are attractive or similar to reference photo
In-door advertisement targeting
In-store and omni-channel identification of returning customers
Media and usability usability
Sorting photo albums
Security
Smart surveillance
Home surveillance
Law enforcement
National Biometric Identification System
Border control
6. Features
• Find position
• Get face keypoints
• Determine
orientation
Face Detection
• Gender
• Age
• Mood
• Hair/Eye/Skin Color
Classification
• Encode face
signature
• Compare to known
persons
• Store and tag
signature
Feature Extraction
& Analysis
17. Technical Performance
Detection, analysis on a 1MP photo and comparison against 5000 reference
profiles (actors)
Time: 2.3-2.5 seconds on one core
Throughput/core/hour: >1000 pictures (assuming 1MP average size)
Scalable, distributed architecture
Executes on several multi-core machines
Highly available data store with replication (Cassandra)