User:Cristinac/Day2: Difference between revisions
No edit summary |
No edit summary |
||
Line 1: | Line 1: | ||
Day four | |||
Deconstructing Harry | |||
Guttorm Guttormsgaard | |||
Asger Jorn | |||
metadata of information; data gallery; average colour of the image, timestamp, face recognition, average colour data, what could a photo gallery mean? | |||
Gaussian blur (many image treatments begin with) | |||
derivates | |||
http://programmingcomputervision.com/ | |||
- | |||
- | a gradient has a magnitude and a direction (like a vector) | ||
Search By Image - Sebastian Schmieg | |||
Control detection | |||
contours that are detected are not continuous, but they are fragments and there is an extra step that determines what kind of fragments got together and create an extra step | |||
Volterra Kernel Training/Identification System | |||
http://www.sciencedirect.com/science/article/pii/S0952197612002461 | |||
statistical, not logical model of the face | |||
behind the algorithm is a manual work that is done by people repeatedly | |||
every detail of the face is annotated | |||
labour conditions | |||
Training data to feed the classifier : no image exists in isolation | |||
False positive : images that have been selected as containing a face when they don’t | |||
https://en.wikipedia.org/wiki/Ghostwriter | |||
cvdazzle.com - techniques to avoid face detection | |||
the same algorithm can be fed with any kind of statistical data; ex: banana recognition | |||
sort by face | |||
http://www.cise.ufl.edu/~arunava/papers/cvpr09.pdf | |||
https://en.wikipedia.org/wiki/Volterra_series | |||
CSV-no space for metadata, no authorship information | |||
https://okfn.org/ | |||
frictionless data | |||
http://centraldedados.pt/ | |||
adding “I think” at the end of every paragraph | |||
iPython |
Revision as of 11:46, 23 January 2015
Day four
Deconstructing Harry Guttorm Guttormsgaard Asger Jorn
metadata of information; data gallery; average colour of the image, timestamp, face recognition, average colour data, what could a photo gallery mean?
Gaussian blur (many image treatments begin with)
derivates
http://programmingcomputervision.com/
a gradient has a magnitude and a direction (like a vector)
Search By Image - Sebastian Schmieg
Control detection contours that are detected are not continuous, but they are fragments and there is an extra step that determines what kind of fragments got together and create an extra step
Volterra Kernel Training/Identification System
http://www.sciencedirect.com/science/article/pii/S0952197612002461
statistical, not logical model of the face
behind the algorithm is a manual work that is done by people repeatedly
every detail of the face is annotated
labour conditions
Training data to feed the classifier : no image exists in isolation False positive : images that have been selected as containing a face when they don’t
https://en.wikipedia.org/wiki/Ghostwriter
cvdazzle.com - techniques to avoid face detection
the same algorithm can be fed with any kind of statistical data; ex: banana recognition sort by face
http://www.cise.ufl.edu/~arunava/papers/cvpr09.pdf
https://en.wikipedia.org/wiki/Volterra_series
CSV-no space for metadata, no authorship information https://okfn.org/
frictionless data http://centraldedados.pt/
adding “I think” at the end of every paragraph
iPython