How to do real-time face tracking?

[storewaladotcom]

Is there a better way to generate a bmp other than the code below:

bufferTexID = Common.genBlankTextureID(mRecordWidth, mRecordHeight);
                    frameBufferObject.bindTexture(bufferTexID);
                    GLES20.glViewport(0, 0, mRecordWidth, mRecordHeight);
                    mFrameRecorder.drawCache();
                    buffer = IntBuffer.allocate(mRecordWidth * mRecordHeight);
                    GLES20.glReadPixels(0, 0, mRecordWidth, mRecordHeight, GLES20.GL_RGBA, GLES20.GL_UNSIGNED_BYTE, buffer);
                    bmp = Bitmap.createBitmap(mRecordWidth, mRecordHeight, Bitmap.Config.ARGB_8888);
                    bmp.copyPixelsFromBuffer(buffer);
                    Log.i(LOG_TAG, String.format("w: %d, h: %d", mRecordWidth, mRecordHeight));

The code is causing a minor lag in the UI. My use case = While the video is recording, take a BMP periodically and check if there is a face in the BMP. If there is another cleaner way of doing this, please suggest.

[storewaladotcom]

any pointers on how to incorporate CGEFaceTracker into CGEFrameRecorder? That may address my issue as well.

[wysaid]

You may need to do this in another thread. Just use CGESharedGLContext to create another context( shared with the view) And put your all code in the background thread.

[wysaid]

The CGEFaceTracker is not so correct when doing real-time processing. I will add a new case for this issue later.

[storewaladotcom]

thank you, I will look forward to that code.

[storewaladotcom]

Could you provide some code on using the sharedcontext please?

[wysaid]

https://github.com/wysaid/android-gpuimage-plus/blob/master/cgeDemo/src/main/java/org/wysaid/cgeDemo/TestCaseActivity.java#L151

[storewaladotcom]

Thank you. I did see that piece of code. how can i get a Bitmap from the shared context?

[wysaid]

You should not use the Bitmap. It's extremely slow. I will do this issue later. And you should change the title to How to do real-time face tracking.

[storewaladotcom]

Thank you @wysaid . I will look forward to this when you get a chance to implement. I went ahead and changed the title.

[storewaladotcom]

@wysaid could you point me to do face recognition at the c++ layer?

[wysaid]

You may take look at the pr I'm very busy recently, and I will do that later.