Parallel Coordinates in Computer Vision - VUT FIT http://www.fit.vutbr.cz/research/groups/graph/pclines/
run:
for compiling mex file for edgelets detection
mex mx_lines
for compiling mex file for diamond space accumulation
mex mx_raster_space
for diamond space accumulation and detection
R = diamond_vanish(InputImg, Normalization, SpaceSize, PatchSize, VanishNumber) input: InputImg - input image Normalization - float number; normalization of the image (1 means normalization form -1 to 1) SpaceSize - int number; resolution of the accumulation space (final space has dims SpaceSize x SpaceSize) PatchSize - int row vector; radius of a patch, from which edge pixels are extracted and use for ellipse fitting for detection of orientation of the edge point VanishNumber - number of vanishing points to be detected output: R - structure with fields R.Space - accumulated diamond space (further is used for orthogonalization) R.PC_VanP - positions of the maxima in R.Space R.PC_VanP_Norm - normalized position of the maxima (R.Space bounded from -1 to 1) R.CC_VanP - position of the vanishing point in the input image coordinates
for evaluation on a dataset
R = run_on_dataset(DB, DB_path, Normalization, SpaceSize, PatchSize) input: DB - cell array with images names (e.g. Manhattan_Image_DB_Names) DB_path - path to dataset data (with camera parameters and image data) Normalization - same as above (if empty, default paramters are used, Normalization = 0.4) SpaceSize - same as above (if empty, default paramters are used, SpaceSize = 321) PatchSize - same as above (if empty, default paramters are used, PatchSize = [6:4:22]) (default parameters are from training on a ECCV_TrainingAndTestImageNumbers -> trainingSetIndex) output: R - structure with fields: R.Image - image name R.Detected - 3x3 matrix with three detected vanishing points, each in one column R.Orthogonal - 3x3 matrix with three vanishing points after orthogonalization, each in one column
#Notes
Manhattan_Image_DB_Names inside YorkUrbanDB / inside Matlab - simply click on Manhattan_Image_DB_Names.mat
Results = run_on_dataset(Manhattan_Image_DB_Names, '/Documents/BMVC-Dubska/YorkUrbanDB')
Known issue -