OULU-NPU: A Mobile Face Presentation Attack Database with Real-World Variations
read more
Citations
Learning Deep Models for Face Anti-Spoofing: Binary or Auxiliary Supervision
Presentation Attack Detection Methods for Face Recognition Systems: A Comprehensive Survey
Searching Central Difference Convolutional Networks for Face Anti-Spoofing
Multi-Adversarial Discriminative Deep Domain Generalization for Face Presentation Attack Detection
Deep Tree Learning for Zero-Shot Face Anti-Spoofing
References
Face Spoof Detection With Image Distortion Analysis
On the effectiveness of local binary patterns in face anti-spoofing
A face antispoofing database with diverse attacks
Face liveness detection from a single image with sparse low rank bilinear discriminative model
Face Spoofing Detection Using Colour Texture Analysis
Related Papers (5)
Frequently Asked Questions (14)
Q2. What is the way to collect the data of genuine subjects?
Assuming that the legitimate users are trying to get authenticated in multiple conditions, it is important to collect the data of genuine subjects in multiple lighting conditions from the usability point of view.
Q3. What is the key issue in face antispoofing and image classification in general?
One of the critical issues in face antispoofing and image classification in general is the generalization across different acquisition devices.
Q4. What is the effect of the camera used for capturing the targeted face sample?
The camera used for capturing the targeted face sample will also lead to imperfect color reproduction compared to the legitimate sample.
Q5. How many subjects were recorded in the OULU-NPU presentation attack detection database?
The OULU-NPU presentation attack detection database includes short video sequences of real access and attackattempts corresponding to 55 subjects (15 female and 40 male).
Q6. What is the price range of the Samsung Galaxy S6 edge?
Considering that the acquisition quality of the embedded (both front and rear) cameras can be expected to be growing generation by generation, the authors selected six smartphones with high-quality front-facing cameras in the price range from e250 to e600 for the data collection:• Samsung Galaxy S6 edge (Phone 1) with 5 MP frontal camera.
Q7. What are the main datasets used for face PAD research?
the MSU Mobile Face Spoof Database (MSUMFSD) [13] and the Replay-Mobile database [5] introduced mobile authentication scenarios to public face PAD benchmark datasets.
Q8. What are the two metrics used in the PAD related literature?
In principle, these two metrics correspond to the False acceptance Rate (FAR) and False Rejection Rate (FRR) commonly used in the PAD related literature.
Q9. What are the main issues of the current public datasets?
While existing publicly available databases still continueto be valuable tools for the community, more challenging datasets are needed to reach the next level and solve some fundamental generalization related problems in face PAD.
Q10. What is the effect of the PAI variation on the generalization performance?
The effect of the PAI variation on the generalization performance is investigated by selecting the spoofing attacks created with different PAIs in the train and test conditions.
Q11. What was the way to capture the video?
During each of the three sessions, a high-resolution photo and video of each user was captured using the back camera of the Samsung Galaxy S6 Edge phone capable of taking 16 MP still images and Full HD videos.
Q12. What is the effect of the PAI on the performance of the baseline method?
The results reported in Table IV show that the variation in the PAI decreases the performance of the baseline method from 7.2% to 14.2% in terms of ACER.
Q13. What was the location of the recordings?
• Session 3: The recordings were taken in a small office where the electronic light was switched on and the windows blinds were up and the windows were located in front of the users.
Q14. What is the effect of the PAI on the APCER?
It is not surprising to notice that illumination variation increases specifically the BPCER, while PAI variation has more significant effect on the APCER.