TITLE: Rethinking the image archive with computational vandalism A presentation LGM 2016, London http://libregraphicsmeeting.org/2016/ proposed by: the Scandanavian Institute for Computational Vandalism http://sicv.activearchives.org as represented by: Nicolas Malevé Michael Murtaugh DESCRIPTION (200 words) Today "computer vision" techniques are pervasive, integrated in smartphone cameras, surveillance systems, and cloud-based image search and tagging. Within the last 15 years, CV engineers increasingly use so-called "visual vocabularies" -- statistically determined clusters of algorithmically determined visual "features" -- to index and order image collections. But what precisely are these features in visual terms? What can visual art practices offer to (and receive from) these techniques? This project is inspired by the Scandinavian Institute of Comparative Vandalism (SICV) founded by Asger Jorn shortly after leaving the Situationist International in 1961. The SICV was an art and research association experimenting with the practice of collage, the forces of photography, image archives, and political imaginaries. Today Computational Vandalism challenges the assumptions at the basis of computer vision, the discipline that programmatically ties together the visual and the mnemonic. Work of the SICV has taken several forms: software "probes" (documented in blog form), installations, and publications. We will present an overview of the work, addressing the questions: *What does computer vision have to offer to creative graphic design practitioners? *When does a cheesy effect become a scientific instrument? *How does an instrument of surveillance become a digital paintbrush? *How is an image more than the sum of it's pixels? Website: http://sicv.activearchives.org/ BIO: Michael Murtaugh (1972, USA) designs and researches community databases, interactive documentary, and tools for new forms of reading and writing online. He teaches on the Masters Degree programme in Media Design and Communication at the Piet Zwart Institute in Rotterdam. Nicolas Malevé (1969, Belgium) is an artist, software programmer and data activist developing multimedia projects and web applications for and with cultural organizations. His current research work is focused on cartography, information structures, metadata, archives and algorithms. ============================== Lecture 17: Bag of words https://youtu.be/iGZpJZhqEME?t=15m55s 16:12 --> ... Descriptor for each interest point. High dimensional space. Cluster them, and this is our vocabulary. English language has a vocabulary [since] 1000s of years, visual language we don't have a vocabulary so we have to come up with our own vocabulary. SICV See slides http://guttormsgaard.activearchives.org/surf/ http://guttormsgaard.activearchives.org/surf/pairs/ Princess and the Pea... / Bag of VIsual Words http://sicv.activearchives.org/logbook/bag-of-visual-words/ Manually annotated words http://sicv.activearchives.org/features/03_annotations.html http://sicv.activearchives.org/features/04_sift_features.html http://sicv.activearchives.org/features/05_collage.html http://sicv.activearchives.org/audio/randomwalk2.webm Vitrines samples with Jorn http://sicv.activearchives.org/vitrine/20160303/collage20160303_163425.webm http://sicv.activearchives.org/vitrine/20160305/collage20160305_131451.webm http://sicv.activearchives.org/vitrine/20160303/collage20160303_224323.webm http://sicv.activearchives.org/vitrine/20160304/collage20160304_220741.webm http://sicv.activearchives.org/vitrine/20160304/collage20160304_182423.webm http://sicv.activearchives.org/vitrine/20160304/collage20160304_220212.webm http://sicv.activearchives.org/vitrine/20160305/collage20160305_011251.webm http://sicv.activearchives.org/vitrine/20160305/collage20160305_013026.webm http://sicv.activearchives.org/vitrine/20160305/collage20160305_111701.webm Feature film: Faces http://sicv.activearchives.org/video/jorn.faces.webm ( WOULD it not be nice to talk about the "3rd" faces ... ie the faces from training ) http://docs.opencv.org/2.4/doc/user_guide/ug_traincascade.html (Showing the contents of the vec file) opencv_createsamples -vec trainingfaces_24-24.vec -w 24 -h 24 https://www.youtube.com/watch?v=eay7CgPlCyo Deep Learning? Are hand crafted features old fashioned? MM todo. check centering of jorn / contours (fit with resoution) Random walk with Jorn would be nice (and faces?) TRANSLATE SLIDES INTO WEB/Leaflet page ... LINKABLE! THIS PAGE: http://pad.constantvzw.org/public_pad/lgm2016-sicv