EyaCV

Quick Revue of Traditional Computer Vision Techniques

color (object localisation / skin region) + morphological filters

• different spaces : RGB, HSV, etc.

object detection (where is Waldo / clef de sol)

• sliding window
• pyramid
• template matching
• non maximum suppression

texture (hortensia Kaggle classification)

• fourrier transform (cosine transform)
• Local Binary Pattern

form (cars detection)

• Histogram of Gradient
• Viola & Jones

point detectors (paper corner finding) + affine transformation

• Harris Corner
• SIFT Detector (SURF faster, HORB open source)
• Paronamic matching

descriptors (naive bag of word boats vs planes image classification)

• SIFT, Horb etc
• scene descriptor : GIST, bag of features
• image retrieval / classification

segmentation (shadow detection)

• color based segmentation
• mean shift, grab cut, CRF

neural network (flower classification)

• classification (VGG, ResNet, Inception, etc.)

More on neural networks in another repository...

• localisation (Yolo)
• segmentation (FB)
• one shot learning (siamese network)
• generation (GAN)