Appendix C. FAQ#
Q: This document contains high-level descriptions and no code examples nor reference. Where can one find them?
A: The complete type and function reference are provided as an interactive web-based documentation; see the doc/fsdk/index.html inside the LUNA SDK package. The examples are located in the /examples folder and "ExamplesGuide.pdf" is located in /doc folder of LUNA SDK package.
Q: Does FaceEngine support multicore / multiprocessor systems?
A: Yes, all internal algorithm implementations are multithreaded by design and take advantage of multi-core systems. The number of threads may be controlled via the configuration file; see configuration manual "ConfigurationGuide.pdf" or comments in the configuration file for details.
Q: What is the state of GPU support?
A: As of version 2.7 the GPU support is implemented for face detection and descriptor extraction algorithms. Starting from version 2.9 GPU implementations are considered stable.
Q: What speedup may be expected from GPUs?
A: Typically GPUs allow accelerating algorithms by the factor of 2-4 times depending on microprocessor architecture and input data.
Q: Are there any official bindings/wrappers for other languages (C#, Java)?
A: No, such bindings are not provided. FaceEngine officially implements C++ API only, bindings to other languages should be created by users themselves. There are tools to automate this process, like, e.g., SWIG.
Q: Does FaceEngine support DBMS systems?
A: No, FaceEngine implements just computer vision algorithms. Users should implement DBMS communication themselves using serialization methods described in section "Serializable object interface" of chapter "Core concepts" and section "Archive interface" of chapter "Core facility".
Q: What image formats does FaceEngine support?
A: FaceEngine does not implement image format encoding functions. If such functions are required, one should use a third-party library, e.g., FreeImage.
FaceEngine functions typically expect image data in the form of uncompressed unencoded pixel data (RGB color 24 bits per pixel or grayscale 8 bits per pixel).
FaceEngine implements convenience functions like RGB -> grayscale and RGB<-> BGR color conversions. The rationale of this design is explained in section "Image type" of chapter "Core concepts".