Doping SVM with HOG + PCA to Double Accuracy скачать в хорошем качестве

Doping SVM with HOG + PCA to Double Accuracy

00:00 Intro 00:14 Background 00:38 Pure SVM 01:16 PCA 01:30 HOG 02:05 Source code on GitHub and Google Colab 02:30 Code run-through Deep Learning and Neural Networks nowadays are the default way of doing Image Classification. Is it possible to achieve good result too but with purely statistical models such as Support Vector Machine? This video demonstrate how a RBF SVM model could improve from 3x% to 62% accuracy with the help of Histogram of Gradients for feature extraction, and grayscaling and Principle Component Analysis for dimensionality reduction. These techniques also reduce training time to merely 5 minutes. Written in Python with SKLearn, and instantly runnable on Google Colab. Source code is available under the MIT licence on GitHub: https://github.com/LemuelKL/CIFAR10-H... Please consider starring the repository as a show of gratitude.