Adding benchmarks?

[njern]

It would be interesting to see benchmarks (esp. on embedded devices like the RbP3 / RbP3+) of both the accuracy and the performance of the detectors (personally I am especially interested in the RealNets face detector/extractor) - this would also hopefully help guide people towards your product :)

[symisc]

If you look closely at the sod_realnet_detect() implementation, you'll see that the detection process is a cheap for(;;) loop. Inside that loop, only pixel intensity comparison is done and O(1) hashtable lookup is performed. The heavy computation is already done in the training phase. On modern architectures, this is highly parallelized and compilers should be able to perform very good optimizations such as loop unrolling (clang is very good at this). On a PI3, the RealNet face detector should run at 12 ~ 20 milliseconds that you can state by yourself. We do not plan to release any public benchmark right now. One last thing to note about the RealNet face detector. This detector is specialized in detecting frontal faces only captured from Webcam or phone frontal camera stream to implement for example Snapchat like filters or face recognition. If you are looking for a scale invariant and more robust face detector, then the CNN face model is the right choice but it is slower than RealNet and RAM hungry. All of them are available at download page.

…

-----Original Message----- From: Niclas Jern <notifications@github.com> To: symisc/sod <sod@noreply.github.com> Cc: Subscribed <subscribed@noreply.github.com> Sent: Thu, 21 Jun 2018 18:20 Subject: [symisc/sod] Adding benchmarks? (#6) It would be interesting to see benchmarks (esp. on embedded devices like the RbP3 / RbP3+) of both the accuracy and the performance of the detectors (personally I am especially interested in the RealNets face detector/extractor) - this would also hopefully help guide people towards your product :)

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: #6