Facial Recognition & Analysis
Facial analysis has received significant attention, especially in the past several years due to a wide range of applications such as public surveillance and security, identity verification in the digital world, modeling techniques in multimedia data management, medical analysis, and human-robot interaction.
At Persistent, we are working on the following facial analysis problems:
- Facial recognition – Match a given facial image with a stored collection of identities and return an identity.
- Facial expression recognition – Identify if a facial image is displaying anger, disgust, fear, neutrality, happiness, sadness, or surprise.
- Gender identification – Identify if the user is male/female from the facial image.
Customer Success Snapshot
- Facial analysis is challenging in presence of covariates such as pose, expression, illumination, aging effect, accessories, and occlusion; These covariates introduce high degree of variations in two images of the same person thereby reducing the performance of the recognition algorithms.
- Building a facial recognition system by using very few images per user, sometimes just one image, was needed.
- Created image datasets for facial expressions as well as face recognition considering the real world environment with poor illumination and occlusion.
- Designed rich features using facial landmarks which help to handle illumination challenge.
- Deep neural networks are trained on large number of images having all possible variations to predict the identity, expression and age of the user; transfer learning is also used.
- The accuracy of the systems is measured on standard public datasets as well as private datasets; we also used real-time video feed from webcams to train the system.
- Face recognition and facial expression recognition systems achieved accuracies of more than 95%.
- The accuracy of the gender identification system was 96%.
You May Be Interested In
Whitepaper: Facial Analysis: Face Recognition, Expression Recognition, and Gender Identification