FPGA face and object detection

[Hofbrau]

About a year ago I worked on an FPGA based face detection system that achieved reasonable performance on cheap hardware.
We reported performance of about 5 frames per second on VGA resolution images, but it could easily achieve 10-15 fps with the addition of pipelining.
Another bottleneck was our memory implementation. We would wait for a full frame to write into memory before processing and manipulating the original image for display.

I was wondering if anyone has worked on or seen compelling results from real time FPGA based face/object detection?
It would be interesting to see how our results compared to the community and not just journal publications.

For reference our project documentation is archived on: (and yes, its dense)
http://www.dejazzer.com/eigenpi/facedetection/E50FinalReport.pdf

And a very brief demo video:
https://www.youtube.com/watch?v=aj4FEovXVXM&feature=youtu.be

[Scrts]

I saw Xylon demo on face detection, but they were using Zynq:


As you can see, they are also doing other detections: eyes, cheeks and lips. Original video claimed to be 1080p60, but captured at 30FPS.

I don't see any reason why you could not do anything beyond 1080p60, you just need more bandwidth with your DRAM.