You are on page 1of 2

Research Gap

The existing face mask recognition focus only in detecting one specific type of face
mask. So the research gap here is to detect different types of face mask using one
image classifier.
 Auto-Machine Learning Services of google is a cost effective machine learning
platform. The process in the machine learning is too long when it pass through
the GPU or Processing units. But there is a cloud or google services that can be
used.
 The only need is raspberry pi to run the program or the model being trained.
 Pre trained model- the pretrained model is well practiced when it comes in
classifying image. The data set will be trained in the model to make the neural
network even smarter to detect what you want to detect.
 The data set will be conducted by the researchers.

The purpose of machine learning is to feed the data to the system and the system will
creates its own rules. Before, we set all the rules in terms of image processing, but the
length of the code is very long, so the path to image recognition is opposite. Provide the
system a data set to determine the rules for that image recognition. And then the output
rules and called models. To perform the rules, use the method or algorithm RCNN, Fast
CNN, CNN to get from the data set the desired model.

Two types of machine learning:


1. Supervised
2. Unsupervised

QUESTIONS

How to integrate or paano gagamitin ang output ng cloud.

We will export the custom model output as TensorFlowLite to embed it to the raspberry pi

STEP BY STEP PROCEDURE OR PROCESS IN IMAGE RECOGNITION:

First we collect gather data a lot of picture based dun sa 1.

Second lalabelan naten, or annotate or ibboox naten yung mga facemask dun sa mga pic specifically.

After non is titrain na sya yung google autoML the parameters will depends on us and the accuracy
depends on how big the dataset is.

After ng training we can see the statistics or the result if my gusto pa ba tayo baguhin regarding sa
accuracy or my mga false positive or negative.
IF wala na, we can export the model to the TensflowLite format na un para maembed sya sa raspberry pi
using TensorflowLite Framework din.

From Collection of data, paano ipapasok sa cloud then from the cloud paano kukunin ang trained data.
And the train data paano ipprocess or gagamitin for image recognition for raspberry pi.

You might also like