Recognition-by-components: a theory of human image understanding.

By Irving Biederman. 1987

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Neuroscience

Citations in the corpus

Local Relation Networks for Image Recognition (2019, background): “Humans have a remarkable ability to “see the infinite world with finite means” [27, 2].”
ALICE: Active Learning with Contrastive Natural Language Explanations (2020, background): “Hierarchical Visual Recognition Categorical hierarchy is inherent in visual recognition (Biederman, 1987; Feng et al., 2019).”
What can the brain teach us about building artificial intelligence? (2016, background): “cepts for illustration. Structural description models represent visual concepts as compositions of parts and relations, which provides a strong inductive bias for constructing models of new concepts (Biederman, 1987; Hummel & Biederman, 1992; Marr & Nishihara, 1978; van den Hengel et al., 20”
Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories (2004, background): “While Weber et al. [5] have demonstrated encouraging results on identifying automatically three categories in a limited image database, in order to reach human performance one would like to see tens of thousands of categories identiﬁed automatically from possibly millions of images [1].”
Designing deep networks for surface normal estimation (2014, background): “The first primitives proposed took the form of lines [27, 36] and volumetric primitives such as geons [1], but these turned out to be too difficult to detect in natural images.”
Neural Prototype Trees for Interpretable Fine-grained Image Recognition (2020, methodology): “Related is the Classification-By-Components network [44] that learns positive, negative and indefinite visual components motivated by the recognition-by-components theory [7] describing how humans recognize objects by segmenting it into multiple components.”
Transductive Multi-View Zero-Shot Learning (2015, background): “Humans can distinguish 30,000 basic object classes [3] and many more subordinate ones (e.g. breeds of dogs).”
Incremental Few-Shot Object Detection (2020, neutral): “In contrast, humans learn new concepts such as object classes incrementally without forgetting previously learned knowledge [14], and often requiring only a few visual examples per class [36, 3].”
Generalisation in humans and deep neural networks (2018, background): “1 Deep neural networks as models of human object recognition The visual recognition of objects by humans in everyday life is rapid and seemingly effortless, as well as largely independent of viewpoint and object orientation [1].”
How does the brain solve visual object recognition? (2012, neutral): “…belies the computational magnitude of this feat: we effortlessly detect and classify objects from among tens of thousands of possibilities (Biederman, 1987) andwe do sowithin a fraction of a second (Potter, 1976; Thorpe et al., 1996), despite the tremendous variation in appearance that…”
SUN Database: Exploring a Large Collection of Scene Categories (2014, methodology): “2 This difference also explains why our count is much higher than (Biederman 1987) estimate of about 1,000 basic-level objects – we included all object words in our count, not just basic-level terms. Caltech 256 Griffin et al. (2007), Tiny Images Torralba et al. (2008) and ImageNet Deng et al.”
Generalizing from a Few Examples (2019, background): “In contrast, human can recognize around 30,000 classes [16], where collecting sufficient images of each class for machines to learn is very laborious.”
Recognition of natural scenes from global properties: Seeing the forest without representing the trees (2008, unknown): “In this view, the meaning of a real-world scene emerges from the identities of a set of objects contained within it, learned through the experience of object cooccurrence and spatial arrangement ([Biederman, 1981], [Biederman, 1987], [De Graef et al., 1990] and [Friedman, 1979]).”
Learning Shape Abstractions by Assembling Volumetric Primitives (2016, methodology): “We obtain volume-image pairs (Vi, Ii) by rendering ShapeNet models with random lighting and background (as suggested in [32]) and train the image based network to mimic the volume based network’s predictions i.e. we train gθ′ to minimize ‖hθ(Vi)−gθ′(Ii)‖2.”
Local Alignments for Fine-Grained Categorization (2014, methodology): “Inspired by the partial object model of Biederman (1987), Farrell et al. (2011) and Zhang et al. (2012) consider the head of a bird as most discriminative, using it to perform recognition.”
Seeing Sets: Representation by Statistical Properties (2001, background): “Studies in object perception often have as their goal the understanding of how individual items are represented (e.g., Biederman, 1987).”
Comparing deep neural networks against humans: object recognition when the signal gets weaker (2017, background): “The visual recognition of objects by humans in everyday life is typically rapid and eﬀortless, as well as largely independent of viewpoint and object orientation (e.g. Biederman, 1987).”