Hardware requirements#
Mobile installations#
Models provided in distribution package and supported devices.
| Neural network | CPU | ARM |
|---|---|---|
| FaceDet_v2_ |
yes | yes |
| FaceDet_v2_ |
yes | yes |
| FaceDet_v2_ |
yes | yes |
| headpose_v3_ |
yes | yes |
| ags_v3_ |
yes | yes |
| eyes_estimation_flwr8_ |
yes | yes |
| eye_status_estimation_ |
yes | yes |
| mask_clf_v3_ |
yes | yes |
| mouth_estimation_v4_ |
yes | yes |
| face_occlusion_v1_ |
yes | yes |
| model_subjective_quality_v1_ |
yes | yes |
| model_subjective_quality_v2_ |
yes | yes |
| glasses_estimation_v2_ |
yes | yes |
| cnn62m_ |
yes | yes |
| oneshot_rgb_liveness_v8_model_ |
yes | yes |
| oneshot_rgb_liveness_v8_model_ |
yes | yes |
| depth_rgb_ |
yes | yes |
| depth_liveness_v2_ |
yes | yes |
| vlTracker_detection_ |
yes | yes |
| vlTracker_template_ |
yes | yes |
| vlTracker_update_ |
yes | yes |
cnn62m_
.plan is provided in complete Android FaceEngine SDK edition only.
CPU requirements#
Supported CPU architectures:
- x86;
- x86_64;
- armeabi-v7a;
- arm64-v8a.
Per-abi libraries are provided for Android.
Memory requirements#
RAM requirements are given for common for mobile platform verification pipeline.
Storage is amount of space specific version of installation takes on device. For Android app Gradle build system strips symbols from all the dynamic libraries when building a release .apk. As the result .so files in your final app archive will occupy (up to 30-60%, depending on platform) less storage space compared to ones found in the distribution.
"Memory requirements"
| Requirements for | Android |
|---|---|
| RAM | 400 MB |
| Storage Full | 350 MB |
| Storage Frontend | 300 MB |
Number of threads on mobile devices#
The description of according settings you can find in "Configuration Guide - Runtime settings". The setting <param name="numThreads" type="Value::Int1" x="-1" /> means that will be taken the maximum number of available threads. This number of threads is equal to according number of available processor cores. We strongly recommend you to follow this recommendation; otherwise, performance can be significantly reduced.