Quote from: WhyS on November 30, 2017, 10:49:21 AM

Quote from: Biometrids on November 29, 2017, 04:46:11 PM

Quote from: sms301 on November 29, 2017, 04:00:27 PM

What type of filter are you guys using? I did similar type of project back in college. Used Kalman Filter which is one of the best filter tech for Image processing and AI. Curious to know what people use these days for Blockchain.

We will be using a lot of different filters, it depends on which part of the face the filter will be working with.

Here is the direct answer from our algorithm developer:

For face detection filters will be skin filter, edges filter like Canny, Sobel, Histogram of Oriented gradient (HOG) or HAAR. For face recognition is much more difficult and we need classifiers and neuron networks. For convolutions networks the filters are choosen by the training software.
In most cases they are 3x3 or 4x4 matrixes
Also we will have filter for eyes, nose, mouth detection, estimating some proportions and so on

I hope this answer your question, or else feel free to ask

Are you pulling data from third parties or how will you ensure that your algorithm works well from the beginning? Or will it take some time for the algorithm to learn and improve with the increasing amount of data from new users?

We will not be pulling any 3rd parties data. The nodes will run the algorithm, and they will learn over time to recognise the right person. So the algorithm starts from scratch, when the person scans their face an put it into their account the first time.