Semantic classes can be either things (objects with a well-defined shape, e.g. car, person) or stuff (amorphous background regions, e.g. grass, sky). While lots of classification and detection works focus on thing classes, less attention has been given to stuff classes. Nonetheless, stuff classes are important as they allow to explain important aspects of an image, including (1) scene type; (2) which thing classes are likely to be present and their location (through contextual reasoning); (3) physical attributes, material types and geometric properties of the scene. To understand stuff and things in context we introduce COCO-Stuff1, which augments all 164K images of the COCO 2017 dataset with pixel-wise annotations for 91 stuff classes. We introduce an efficient stuff annotation protocol based on superpixels, which leverages the original thing annotations. We quantify the speed versus quality trade-off of our protocol and explore the relation between annotation time and boundary complexity. Furthermore, we use COCO-Stuff to analyze: (a) the importance of stuff and thing classes in terms of their surface cover and how frequently they are mentioned in image captions; (b) the spatial relations between stuff and things, highlighting the rich contextual relations that make our dataset unique; (c) the performance of a modern semantic segmentation method on stuff and thing classes, and whether stuff is easier to segment than things.
The most deliberate part of first-language teaching is the business of telling a child what each thing is called. We ordinarily speak of the name of a thing as if there were just one, but in fact, of course, every referent has many names. The dime in my pocket is not only a dime. It is also money, a metal object, a thing, and, moving to subordinates, it is a 1952 dime, in fact a particular 1952 dime with a unique pattern of scratches, discolorations, and smooth places. When such an object is named for a very young child how is it called? It may be named money or dime but probably not metal object, thing, 1952 dime, or particular 1952 dime. The dog out on the lawn is not only a dog but is also a boxer, a quadruped, an animate being; it is the landlord's dog, named Prince. How will it be identified for a child? Sometimes it will be called a dog, sometimes Prince, less often a boxer, and almost never a quadruped, or animate being. Listening to many adults name things for many children, I find that their choices are quite uniform and that I can anticipate them from my own inclinations. How are these choices determined and what are their consequences for the cognitive development of the child? Adults have notions about the kind of language appropriate for use with children. Especially strong and universal is the belief that children have trouble pronouncing long names and so should always be given the shortest possible names. A word is preferable to a phrase and, among words, a monosyllable is better than a polysyllable. This predicts the preference for dog and Prince over boxer, quadruped, and animate being. It predicts the choice of dime over metal object and particular 1952 dime. Zipf (10) has shown that the length of a word (in phonemes or syllables) is inversely related to its frequency in the printed language. Consequently the shorter names for any thing will usually also be the most frequently used names for that thing, and so it would seem that the choice of a name is usually predictable from either frequency or brevity. The monosyllables dog and Prince have much higher frequencies according to the Thorndike-Lorge list (8) than do the polysyllables boxer, quadruped, and animate being. It sometimes happens, however, that the frequency-brevity principle makes the wrong prediction. The thing called a pineapple is also fruit. Fruit is the shorter and more frequent term, but adults will name the thing pineapple. Similarly they will say apple, banana, orange, and even pomegranate; all of them longer and less frequent words than the perfectly appropriate fruit. Brevity seems not to be the powerful determinant we had imagined. The frequency principle can survive this kind of example, but only if it is separated from counts like the Thorndike-Lorge of over-all frequency in the printed language. On the whole the word fruit appears more often than the word pineapple (and also is shorter), but we may confidently assume that, when pineapples are being named, the word pineapple is more frequent than the word fruit. This, of course, is a kind of frequency more directly relevant to our problem. Word counts of general usage are only very roughly applicable to the prediction of what will be said when
Scientists have uncovered a new explanation for what powers Yellowstone and other supervolcanoes。 Instead of a deep plume rising from near Earth’s core, a broad “mantle wind” may push hot rock beneath Yellowstone, generating magma closer to the surface。 This process helps create a massive underground magma network and may explain how supervolcanoes
The battle between butter and margarine comes down to chemistry。 Butter’s naturally occurring fats create rich flavors, golden browning, and flaky baked goods, while margarine’s modified plant oils offer consistency and a longer shelf life。 Although both contain mostly fat, their different structures influence how they behave in recipes
Three meanings of “information” are distinguished: “Information-as-process”; “information-as-knowledge”; and “information-as-thing,” the attributive use of “information” to denote things regarded as informative. The nature and characteristics of “information-as-thing” are discussed, using an indirect approach (“What things are informative?”). Varieties of “information-as-thing” include data, text, documents, objects, and events. On this view “information” includes but extends beyond communication. Whatever information storage and retrieval systems store and retrieve is necessarily “information-as-thing.” These three meanings of “information,” along with “information processing,” offer a basis for classifying disparate information-related activities (e.g., rhetoric, bibliographic retrieval, statistical analysis) and, thereby, suggest a topography for “information science.” © 1991 John Wiley & Sons, Inc.
List of Illustrations and Tables Preface 1. Instrument Epistemology 2. Models: Representing Things 3. Working Knowledge 4. Encapsulating Knowledge 5. The Instrumentation Revolution 6. Thing Knowledge 7. The Thing-y-ness of Things 8. Between Technology and Science 9. Instrumental Objectivity 10. The Gift References Index
Humanity has sat at the center of philosophical thinking for too long. The recent advent of environmental philosophy and posthuman studies has widened our scope of inquiry to include ecosystems, animals, and artificial intelligence. Yet the vast majority of the stuff in our universe, and even in our lives, remains beyond serious philosophical concern. This book develops an object-oriented ontology that puts things at the center of being—a philosophy in which nothing exists any more or less than anything else, in which humans are elements but not the sole or even primary elements of philosophical interest. And unlike experimental phenomenology or the philosophy of technology, this book’s alien phenomenology takes for granted that all beings interact with and perceive one another. This experience, however, withdraws from human comprehension and becomes accessible only through a speculative philosophy based on metaphor.
To paraphrase Cole Porter's famous 1926 song, "What is this thing called pain? This funny thing called pain, just who can solve its mystery?" Pain, like love, is all consuming: when you have it, not much else matters, and there is nothing you can do about it. Unlike love, however, we are actually beginning to tease apart the mystery of pain. The substantial progress made over the last decade in revealing the genes, molecules, cells, and circuits that determine the sensation of pain offers new opportunities to manage it, as revealed in this Review series by some of the foremost experts in the field.
this paper to show that this counter-response cannot be sustained and that the original argument against identity should therefore be allowed to stand. This is no easy task since there would appear to be nothing in the immediate linguistic data to settle the question one way or the other. However, by working through the consequences of the counter-response for the rest of our language, I think it may be shown to be extremely implausible. The paper is in two main parts. The first (sections --) is largely concerned with setting up the problem. We characterize the dierent forms the identity theory can take (section ), explain how the argument in favour of non-identity might in principle break down (section ), present the most plausible versions of such arguments (section ), and then consider the most plausible counter-response to them (section ). The second part (sections --) embarks on a detailed investigation of the diculties with the counter-response. It is shown to be unable to account for a wide variety of dierent linguistic data, that is loosely classified according as to how reference to a material thing might be achieved. Four main kinds of case will be considered: those in which a sort is explicitly invoked (section ); those in which it is implicitly invoked (section \t); those in which the very notion of reference is itself used in securing reference (section ); and those in which there is reference to a plurality of things (section )
What if our biggest idea about reality is built on a hidden misunderstanding。 A new philosophical look at space-time challenges the popular view that the past, present, and future all exist together in a timeless "block universe。" The argument suggests that physicists may be blurring the difference between things that exist and things that merely o
It’s not often you can score discounts from the outdoor-coded Canadian company that makes understated and stylish performance clothing
A lightweight new X-ray telescope could finally give scientists something they’ve never had before: a complete chemical map of the Moon。 Researchers used detailed mission simulations to show that a compact telescope orbiting the Moon could identify key elements across the entire lunar surface, helping reveal how the Moon formed and evolved
A new sunlight-powered material can convert visible light into higher-energy UV light, overcoming a challenge that has frustrated scientists for years。 The breakthrough could enable cleaner air purification, solar-driven chemistry, and advanced manufacturing technologies using nothing more than natural sunlight
Astronomers may have witnessed one of the rarest and most dramatic cosmic events ever seen: a long-sought intermediate-mass black hole ripping apart a dense white dwarf star and devouring it。 The Einstein Probe space telescope caught the explosion in its earliest moments, revealing an unusual sequence of intense X-ray flashes unlike anything seen i
SETI scientists searched the interstellar object 3I/ATLAS for radio signals that could indicate extraterrestrial technology but found nothing beyond human-made interference。 Even so, the rapid-response observations helped confirm the object's natural origin and showcased how future interstellar visitors can be investigated for signs of intelligent
Astronomers have finally cracked the mystery of the famous “Pink Planet,” a strange world 57 light-years away that has puzzled scientists for more than a decade。 Using the James Webb Space Telescope, researchers discovered that its atmosphere contains water vapor, methane, carbon dioxide, ammonia, and something never directly confirmed before in su
Scientists found that transfer learning can make the search for new physics in the universe much faster, slashing the need for expensive simulations。 Yet the approach can backfire when AI relies too heavily on familiar patterns, potentially missing evidence of something truly new