2018/11/11 by Ahmed El-Sayed, Elif Kongar, ElSayed, Ahmed +8
Computer Science · Engineering · #3D Shape Modeling and Analysis #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Face and Expression Recognition #Face recognition and analysis
paper · pdf · doi:10.48550/arxiv.1811.04358
openalex publication_date 2018/11/11 · openalex created_date 2022/08/02 · openalex updated_date 2026/07/28
Using heterogeneous depth cameras and 3D scanners in 3D face verification\ncauses variations in the resolution of the 3D point clouds. To solve this\nissue, previous studies use 3D registration techniques. Out of these proposed\ntechniques, detecting points of correspondence is proven to be an efficient\nmethod given that the data belongs to the same individual. However, if the data\nbelongs to different persons, the registration algorithms can convert the 3D\npoint cloud of one person to another, destroying the distinguishing features\nbetween the two point clouds. Another issue regarding the storage size of the\npoint clouds. That is, if the captured depth image contains around 50 thousand\npoints in the cloud for a single pose for one individual, then the storage size\nof the entire dataset will be in order of giga if not tera bytes. With these\nmotivations, this work introduces a new technique for 3D point clouds\ngeneration using a neural modeling system to handle the differences caused by\nheterogeneous depth cameras, and to generate a new face canonical compact\nrepresentation. The proposed system reduces the stored 3D dataset size, and if\nrequired, provides an accurate dataset regeneration. Furthermore, the system\ngenerates neural models for all gallery point clouds and stores these models to\nrepresent the faces in the recognition or verification processes. For the probe\ncloud to be verified, a new model is generated specifically for that particular\ncloud and is matched against pre-stored gallery model presentations to identify\nthe query cloud. This work also introduces the utilization of Siamese deep\nneural network in 3D face verification using generated model representations as\nraw data for the deep network, and shows that the accuracy of the trained\nnetwork is comparable all published results on Bosphorus dataset.\n