Skip to content

Computer Vision multilabel classification with different techniques: (KNN,PCA,LDA,BOVW,SIFT,VGG16)

Notifications You must be signed in to change notification settings

StavC/MMN12ComputerVisionUniversity

Repository files navigation

MMN12ComputerVisionUniversity

Computer Vision multilabel classification with different techniques: (KNN,PCA,LDA,BOVW,SIFT,VGG16)

The full Report can be found at the Doc file.

Question 1:

Classify categories of MNIST Fashion by using KNN,PCA, LDA:

alt text

KNN Accuracy (takes alot of time to fit)

alt text

PCA with 20 Main Components(Good accuracy and fast fitting)

alt text

LDA with 8 componenets(fast fitting and ok accuracy)

alt text

transforming picture to lower dimension with PCA and inversing back

alt text

Question 2:

multilabel classification with SIFT or VGG16 as Feature extractor, Bag Of Visual Words representation and classification with LinearSVM.

DataBase: http://people.csail.mit.edu/torralba/code/spatialenvelope/spatial_envelope_256x256_static_8outdoorcategories.zip

alt text

Sift Method

alt text

VGG Method + ROC + Confusion Matrix

alt text alt text alt text

About

Computer Vision multilabel classification with different techniques: (KNN,PCA,LDA,BOVW,SIFT,VGG16)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages