Working notes from the Scandinavian Institute for Computational Vandalism

Tag Archives: training data

The cloudy days of machine learning

Once upon a time, the US Army wanted to use neural networks to automatically detect camouflaged enemy tanks. The researchers trained a neural net on 50 photos of camouflaged tanks in trees, and 50 photos of trees without tanks. Using standard techniques for supervised learning, the researchers trained the neural network to a weighting that […]

Convolutional Network, 1993

In 1988, Yann LeCun joined the Adaptive Systems Research Department at AT&T Bell Laboratories in Holmdel, New Jersey, United States, headed by Lawrence D. Jackel, where he developed a number of new machine learning methods, such as a biologically inspired model of image recognition called Convolutional Neural Networks, the “Optimal Brain Damage” regularization methods, and […]

Comparative Ways of Seeing

Your browser does not support the video tag. In the frame of Ways of Machine Seeing, a series of experiments in collaboration wit Geoff Cox on the the four episodes of the BBC documentary series Ways of Seeing. In this probe, the same algorithm runs an object detection script using two different training sets. For […]

Native contours

In this experiment [Hochberg & Brooks, 1962], a human baby was raised until the age of 19 months under the constant supervision of his parents who avoided exposing the child to line-drawings or two-dimensional pictures of any kind. Although the baby accidentally had opportunities to glance at some pictures on a few occasions, at no […]

Detecting the Manchurian Candidate

In the Manchurian Candidate, it is revealed that the Communists have been using the soldier Shaw as a sleeper agent, a guiltless assassin subconsciously activated by seeing the “Queen of Diamonds” playing card while playing solitaire. Provoked by the appearance of the card, he obeys orders which he then forgets. Isn’t an image just like […]

Green Mamba

Tensorflow, an open source library for Machine Learning developed by Google, comes with a demo program that labels images. The output of the program contains the labels (or class names) and their scores. The score represents the probability of the image to belong to a certain class. green mamba (score = 0.43074) vine snake (score […]

250 000 labels

This document, edited by Antonio Torralba, contains the notes written by Adela Barriuso describing her experience while using the LabelMe annotation tool. Mrs Barriuso has no training in computer vision. In 2007 she started to use LabelMe to systematically annotate the SUN database. The goal was to build a large database of images with all […]

Capta

Etymologically the word data is derived from the Latin dare, meaning ‘to give’. In this sense, data are raw elements that can be abstracted from (given by) phenomena – measured and recorded in various ways. However, in general use, data refer to those elements that are taken; extracted through observations, computations, experiments, and record keeping […]

The Annotator

A report from one of the Cqrrelations working groups: From the start we were interested in how a Gold Standard is established, a paradoxical situation where human input is both considered a source of truth, and made invisible. Annotation here means the manual work of ‘scoring’ large amounts of data that can than be used […]