Dear Joseph, please have a look the paper in attachment, at the first.

All the best.

Could you clarify your question Dear Joseph, because you talk about Face detection and face Recognition and those are different problems with different results. Face detection can be accomplished very well using Viola - Jones algorithm. Generally, but depending on the face recognition approach, you would need to make training using face images at least in the appearance based approach, and this is face dependent, so race dependent. 

Dear Joseph Mwema ,

Read my article, may be useful.

https://www.researchgate.net/publication/279299807_System_of_recognition_of_an_accessory_of_persons_to_racial_and_an_ethnic_group_on_the_basis_of_photographs

Regards, Shafagat

Article System of recognition of an accessory of persons to racial a...

Real advices and good help.

According to a fairly recent study,

Phillips.IVC2014.Comparison of human and computer performance across face,

both human and machine vision systems exhibit the other-race effect in facial recognition. Take a look at Section 4 of the mentioned paper for more details:

"The study showed an other-race effect for the algorithms. Specifically, performance of face recognition algorithms varies as a function of the demographic origin of the algorithm and the demographic contents of the test population. The mechanisms underlying the other-race effect for humans are reasonably well understood and are based in early experience with faces of different races. Because the algorithms tested were black boxes, conclusions about the mechanisms underlying the algorithm effects are tentative, but the effects reported were not. The results point to an important performance variable combination that has not received much attention. The results also suggest a need to understand how the ethnic composition of a training set impacts the robustness of algorithm performance. From a practical perspective, algorithms need to be evaluated on face sets whose demographics match those at the locations where they will be used."

http://www.nist.gov/manuscript-publication-search.cfm?pub_id=913011

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=6395125&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D6395125

Dear Eugene! Very professional answer many questions. All the best. Vladimir

There are out there several free-to-download face databases such as the BioID Face Database, the FERET Database, the Yale Face Database, etc. They encompass multiple images per subject (multiracial sometimes), per different facial expression, per poses and per illumination conditions. If you extend your constructed [East African] model to include one of those databases, I believe it could take away any doubts you may have about this issue.

Thank you all for your insightful contribution. I am so grateful to each and everyone of you. Your contributions and shared publications on the same are highly appreciated and have given me an informed perspective in this area of research.