Raspberry Pi: The low training Santa/Not Santa detector

64 views Leave a comment

Did we see Mommy kissing Santa Claus? Or was it simply an imposter? The Not Santa detector is here to assistance solve a poser once and for all.

The video is a demo of my “Not Santa” detector that we deployed to a Raspberry Pi. we lerned a detector regulating low learning, Keras, and Python. You can find a full source formula and educational here: https://www.pyimagesearch.com/2017/12/18/keras-deep-learning-raspberry-pi/


Ho-ho-how does it work?
Note: Adrian Rosebrock is not Santa. But he does a good adequate sense of a ridicule aged associate that his costume can fool a Raspberry Pi into meditative otherwise.

Raspberry Pi 'Not Santa' detector

We jest, though has anyone seen Adrian and Santa in a same room together?
Image c/o Adrian Rosebrock

But how is a Raspberry Pi means to detect a Santa-ness or Not-Santa-ness of people who travel into a frame?


Two words: low learning
If you’re not certain what low training is, you’re not alone. It’s a vast topic, and one that Adrian has created a book about, so we grilled him for a bluffers’ guide. In his words, low training is:

…a subfield of appurtenance learning, that is, in spin a subfield of synthetic comprehension (AI). While AI embodies a large, different set of techniques and algorithms associated to involuntary logic (inference, planning, heuristics, etc), a appurtenance training subfields are specifically interested in pattern recognition and learning from data.

Artificial Neural Networks (ANNs) are a category of appurtenance training algorithms that can learn from data. We have been regulating ANNs successfully for over 60 years, though something special happened in a past 5 years — (1) we’ve been means to amass vast datasets, orders of bulk incomparable than prior datasets, and (2) we have entrance to specialized hardware to sight networks faster (i.e., GPUs).

Given these vast datasets and specialized hardware, deeper neural networks can be trained, heading to a tenure “deep learning”.

So now we have a bird’s-eye perspective of low learning, how does a detector detect?


Cameras and twinkly lights
Adrian used a indication he had lerned on dual datasets to detect either or not an picture contains Santa. He deployed a Not Santa detector formula to a Raspberry Pi, afterwards trustworthy a camera, speakers, and The Pi Hut’s 3D Christmas Tree.

Raspberry Pi 'Not Santa' detector

Components for Santa detection
Image c/o Adrian Rosebrock

The camera captures footage of Santa in a wild, while a Christmas tree appendage provides a twinkly notification, accompanied by a resonant ho, ho, ho from a speakers.


A deeper low dive into low learning
A full relapse of a plan and the workings of a Not Santa detector can be found on Adrian’s blog, PyImageSearch, that includes links to other low training and picture sequence tutorials regulating TensorFlow and Keras. It’s an glorious place to start if you’d like to know some-more about low learning.


Build your possess Santa detector
Santa competence locate on to Adrian’s crafty detector and start avoiding a camera, and for that eventuality, we have a possess Santa detector. It uses suit showing to forewarn we of his participation (and your presents!).

Raspberry Pi Santa detector

Check out a Santa Detector apparatus here and use a pacifist infrared sensor, Raspberry Pi, and Scratch to locate a large male in action.

Source: Raspberry Pi blog

Comment this news or article