Hi, great video, i'm wondering how can you make it so when you have both hands down it prints none instead of left or right. Could you please make a video on that?
Hello sofia mendivelso guerrero , I used both and that's the beauty of this project , application of ML & DL together 😊 Just look at the "Code for extracting features and train the knn model" (attached in the description box) , there you will notice a line -- featuresTrain = activations(net,augimdsTrain,featureLayer,'OutputAs','rows'); --- here DL is used to extract features And then the next line is -- classifier=fitcknn(featuresTrain,trainingLabels,'NumNeighbors',5,'Standardize',1); --here fitcknn is to train knn model using those extracted feature ... if you think why even we need this , then here is the explanation -- Advantage of Deep Learning -- Feature engineering is not needed in Deep Learning because important features are automatically detected by neural networks. Disadvantage of Deep Learning -- Generally Millions of data points (Big Data) needed to train Deep Learning model for good accuracy , don't work well with small dataset. Advantage of Machine Learning -- A good Machine learning model can be trained on few thousands of data points also , not needed big data always . Disadvantage of Machine Learning -- More human intervention is involved for Feature Extraction step as ML models can't extract the features by itself. So what if we combine both Deep Learning and Machine Learning to overcome their disadvantages ?!?😊 That's what is done in the video ... Alexnet(convolutional neural network(CNN)) is used for Automatic Feature Extraction and then I used those features to train K-Nearest Neighbor model , so that way , we no need to manually extract feature as Alexnet DL is automatically doing that and although the dataset is small but still the accuracy will be pretty good as we are using ML to train and predict the testdata and not DL model... Here are some more projects where I demonstrated same pattern with other Algorithms -- Feature extraction using Alexnet and use SVM for Image Classification th-cam.com/video/mYF2GH-JMWs/w-d-xo.html Featur extraction using Resnet-18 & use kmeans to train ML model for Image segregation th-cam.com/video/7Hv8BY4RyGk/w-d-xo.html Hope this will be helpful ! Happy Learning
Code to create Database:
clc
clear all
close all
warning off;
cao=webcam;
c=200;
temp=0;
while temp
Hi, great video, i'm wondering how can you make it so when you have both hands down it prints none instead of left or right. Could you please make a video on that?
words are segmented and these are different sizes. how to read these images in CNN model in matlab 2017a
in the training.m, what does alexnet do? mine didn't work properly
Alexnet is the pretrained network , we are applying transfer learning on this.
Thanks so much sire
Most welcome! Happy Learning :-)
Why are you saying you're using KNN when actually you're using the pre-downloaded alexnet. I mean, you're doing deep learning
Hello sofia mendivelso guerrero , I used both and that's the beauty of this project , application of ML & DL together 😊
Just look at the "Code for extracting features and train the knn model" (attached in the description box) , there you will notice a line --
featuresTrain = activations(net,augimdsTrain,featureLayer,'OutputAs','rows'); --- here DL is used to extract features
And then the next line is -- classifier=fitcknn(featuresTrain,trainingLabels,'NumNeighbors',5,'Standardize',1); --here fitcknn is to train knn model using those extracted feature ...
if you think why even we need this , then here is the explanation --
Advantage of Deep Learning --
Feature engineering is not needed in Deep Learning because important features are automatically detected by neural networks.
Disadvantage of Deep Learning --
Generally Millions of data points (Big Data) needed to train Deep Learning model for good accuracy , don't work well with small dataset.
Advantage of Machine Learning --
A good Machine learning model can be trained on few thousands of data points also , not needed big data always .
Disadvantage of Machine Learning --
More human intervention is involved for Feature Extraction step as ML models can't extract the features by itself.
So what if we combine both Deep Learning and Machine Learning to overcome their disadvantages ?!?😊
That's what is done in the video ... Alexnet(convolutional neural network(CNN)) is used for Automatic Feature Extraction and then I used those features to train K-Nearest Neighbor model , so that way , we no need to manually extract feature as Alexnet DL is automatically doing that and although the dataset is small but still the accuracy will be pretty good as we are using ML to train and predict the testdata and not DL model...
Here are some more projects where I demonstrated same pattern with other Algorithms --
Feature extraction using Alexnet and use SVM for Image Classification
th-cam.com/video/mYF2GH-JMWs/w-d-xo.html
Featur extraction using Resnet-18 & use kmeans to train ML model for Image segregation
th-cam.com/video/7Hv8BY4RyGk/w-d-xo.html
Hope this will be helpful ! Happy Learning