I have concerns about this being implemented in the browser process on Android. The FaceDetector class has a Java interface, but it really is just a wrapper around a JNI/native implementation (specifically the neven library). What is the reason for running this code in the browser on data from the renderer? Is there a design doc for this feature? I raised a similar issue about processing renderer-controlled data in the browser on the Mac implementation code review.
Comment 1 by mcasas@chromium.org
, Sep 12 2016