Project for 2019 Skoltech Intro to CV course
Angelina Yaroshenko, Dinar Sharafutdinov, Dmitry Vypiraylenko
Implementation of pedestrian and traffic signs detection using HOG + SVM with non-maxima suppression. We compared results with pretrained Faster R-CNN. Additionally, we trained a convolution neural net for traffic signs classification. As a training dataset, we've used Oscar dataset . The specific thing about data is that it's winter pictures from the car. For positive samples, we've used Russian Traffic Sign Dataset, Caltech Pedestrian Detection Benchmark, and German Traffic Sign Recognition Benchmark
Results
Detection non-maxima suppression (without/with)
Detection mistakes
Final traffic signs detection
Final pedestrian detection
Faster R-CNN for pedestrian detection
