Data Set :　Personal Photos classification
Classmates photos (Original Resolution)(Training and Testing Datasets)
DeepLearning_ClassmatesImageClassification_jdwang2018_6_1.7z
python program (ClassmatesPhotosClassification_YourOwnDataset-TrainAndTest.ipynb)
photos (data/traing, validation, test)
Prepare your own personal photos for Training, validation and testing dataset.
ALL (32,64,128,256(New),512(New))
jdwang (class:1)
Karamo(class:2)
Bambang(class:3)
Bruk(class:4)
Efendi(class:5)
Markus(class:6)
Mikko(class:7)
Image Resolution Reduction
Python Code Example for image resolution reduction
What kind of experiments you have tried to improve the classification accuracy? Did they did work? Why or Why not?
Image resolution? higher : 256, 512, 1024,...
Increasing the size of training data? (each of classmates provides more personal photos. e.g. 20, 30, 50, 100 photos)
The way of taking photos? (face-images from various views? distance? clothes? or background noisy ? )
Deep Learning Models? (MLP? CNN?) (more epochs?) (more layer?)(activation function?)
What is the best accuracy you achieved?
What are the points you have learned from yours experiements that can improve these classmates personal-images classification?