Finding Pictures of Objects in Large Collections of Images

By D.A. Forsyth, Jitendra Malik, Margaret M. Fleck, Hayit Greenspan, Thomas K. Leung, Serge J. Belongie, Chad Carson, C. Bregler. 1996

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

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

In Computer Vision and Pattern Recognition

Citations in the corpus

COCO-Stuff: Thing and Stuff Classes in Context (2016, background): “(2) Size: Things occur at characteristic sizes with little variance, whereas stuff regions are highly variable in size [21, 2, 27].”
Efficient Piecewise Training of Deep Structured Models for Semantic Segmentation (2015, background): “nt types of context. In the PASCAL VOC dataset [10], the image regions of 20 object categories are labeled in the training data, and all remaining regions are labeled as “background”. As discussed in [12], [17], the objects and “background” respectively correspond to the notion of “things” and “stuf”
Weakly- and Semi-Supervised Panoptic Segmentation (2018, background): ““Stuff” classes (such as “sky” and “vegetation”), on the other hand, are amorphous regions of homogeneous or repetitive textures [14].”
Panoptic Segmentation with a Joint Semantic and Instance Segmentation Network (2018, background): “At a high level these elements can be divided into two categories: things and stuff [4].”
On the Robustness of Semantic Segmentation Models to Adversarial Attacks (2017, background): “Cityscapes has, subjectively, less variability than VOC and it also labels “stuff” classes [27].”
BING: Binarized normed gradients for objectness estimation at 300fps (2014, unknown): “Objects are stand-alone things with well-deﬁned closed boundaries and centers [4, 31, 40] although the visibility of these boundaries depends on the characteristics of the background of occluding foreground objects.”
Multi-class cosegmentation (2012, background): “but might also be a texture (grass, rocks), or other “stuff” (a building, a forest) [2].”
Contextualizing Object Detection and Classification (2015, background): “In their work, the terms “stuff” and “things” (originally introduced by Forsyth et al. [19]) are used to distinguish “material” that is deﬁned by a homogeneous or repetitive pattern of ﬁne-scale properties, but has no speciﬁc or distinctive spatial extent or shape (stuff) from “objects with speciﬁc…”
Towards Bounding-Box Free Panoptic Segmentation (2020, background): “[10] of splitting the objects categories into things (countable objects like persons, cars, etc.”
Semantic Amodal Segmentation (2015, methodology): “Using common terminology [1, 11], we explicitly classify the labels into two categories: ‘things’ and ‘stuff’, where a ‘thing’ is an object with a canonical shape (person, ﬁsh, ﬂower) while ‘stuff’ has a consistent visual appearance but can be of arbitrary spatial extent (grass, cloud, water).”
Learning Spatial Context: Using Stuff to Find Things (2008, unknown): “[4] as:”
What, Where and How Many? Combining Object Detectors and CRFs (2010, background): “The distinction between these two sets of classes — referred to as things and stuff respectively — is well known [10,11,12].”