3DiVi Face SDK Technical Specifications
3DiVi Face SDK Accuracy (NIST Standards)
The table below shows the 3DiVi Face SDK quality metrics based on the National Institute of Standards and Technology (NIST) Standards:
| NIST Face Recognition Vendor Test (FRVT) 1:1 | Score |
| VISA* True Acceptance Rate (@FAR 1E-6) | 98.49% |
| Mugshot* True Acceptance Rate (@FAR 1E-5) | 99.22% |
| Border* True Acceptance Rate (@FAR: 1E-6) | 98.65% |
| WILD* True Acceptance Rate (@FAR 1E-4) | 96.49% |
Click here to see the description of image types from the table*
- VISA – Full Frontal image type. The images are of size 252x300 pixels. The mean interocular distance (IOD) is 69 pixels.
- Mugshot – Full Frontal image type. The images are of variable sizes. The mean IOD is 113 pixels.
- Border – the images are taken with a camera oriented by an attendant toward a cooperating subject. This is done under time constraints so there are role, pitch and yaw angle variations. Also background illumination is sometimes strong, so the face is under-exposed. There is some perspective distortion due to close range images. Some faces are partially cropped.
- Wild – the images include many photojournalism-style images. Resolution varies very widely. The images are very unconstrained, with wide yaw and pitch pose variation. Faces can be occluded, including hair and hands.
Accuracy of 3DiVi Face SDK Identification Methods
| FAR | 10.30 TAR (%) | 10.100 TAR (%) | 9.300 TAR (%) | 10.1000 TAR (%) |
| 1E-04 | 98.0 | 99.3 | 99.5 | 99.7 |
| 1E-05 | 98.0 | 99.1 | 99.0 | 99.7 |
| 1E-06 | 95.7 | 98.3 | 98.0 | 99.5 |
Accuracy of Face Attributes Detection
| Detection Attributes | Accuracy Rate |
| Gender | 95% |
| Emotion | 80% |
| Age | 3.95 y/o (mean Average Error) |
The tables in this section show the speed measurements for the Face SDK modules using CPU and GPU. For more detailed specifications, see the Performance Parameters section.
Detectors Timing Characteristics*
| Detector | GPU (NVIDIA GTX 1080 Ti) | CPU (Core i7-7700K 4.4GHz) |
| ULD** | 35ms | 200ms |
| BLF** | 14ms | 38ms |
| REFA | 235ms | 518ms |
* Speed is indicated for 4 faces in an image with a resolution of 1280x720.
** More efficient configurations of detectors are available, see the Performance Parameters section for details.
Identification Timing Characteristics
Desktop
| Method | GPU (NVIDIA GTX 1080 Ti) | CPU (Core i7-7700K 4.4GHz) |
| 10v1000 | 16 ms | 690 ms |
| 9v300 | 10 ms | 260 ms |
| 10v100 | 13 ms | 40 ms |
| 10v30 | 11 ms | 24 ms |
Mobile
| Method | CPU (Qualcomm Snapdragon 845) | GPU (Adreno 630) |
| 9v1000 | 3660ms | 610ms |
| 9v300 | 1960ms | 280ms |
| 9v30 | 170ms | 70ms |
Note: Google Pixel 3 was used for the speed test.