120 Matching Annotations
  1. Last 7 days
  2. Aug 2024
    1. For approximate reference, sRGB/Rec709 cover roughly 35% of visible colour. DCI-P3/Display-P3/AdobeRGB cover roughly 50%, and Rec2020/2100 covers roughly 75%. So even "HDR" standards don't cover everything a human being is capable of seeing.
  3. Jun 2024
  4. Apr 2024
  5. Mar 2024
  6. Feb 2024
  7. Jan 2024
  8. Dec 2023
  9. Nov 2023
  10. Oct 2023
    1. (Weight W1) (Rope Rp) (Rope Rq) (Pulley Pa) (hangs W1 from Rp) (pulley-system Rp Pa Rq) (Weight W2) (hangs W2 from Rq) (Rope Rx) (Pulley Pb) (Rope Ry) (Pulley Pc) (Rope Rz) (Rope Rt) (Rope Rs) (Ceiling c) (hangs Pa from Rx) (pulley-system Rx Pb Ry) (pulley-system Ry Pc Rz) (hangs Pb from Rt) (hangs Rt from c) (hangs Rx from c) (hangs Rs from Pc) (hangs W2 from Rs) (value W1 1) (b) P1. P2. P3. P4. .. Single-string support. (weight < Wx>) (rope <Ry >) (value <Wx> <n>) (hangs <Wx> <Ry>) -(hangs <Wx> <Rx>) - (value <Ry> <W-number>) Ropes over pulley. (pulley <P>) (rope <R1>) (rope <R2>) (pulley-system <R1 > <P> <R2>) (value <R1> <nl>) - (value <R2> <nl>) Rope hangs from or supports pulley. (pulley <R1>) (rope <R1>) (rope R2>) (pulley-system <R1> <P> <R2>) { (hangs <R3> from <P>) or (hangs <P> from <R3>) } (value <R1> <nl>) (value <R2> <n2>) - (value <R3> <nl + <n2>) Weight and multiple supporting ropes. (weight <W1 >) (rope <R1 >) (rope R2>) (hangs <W1> <Rl>) (hangs <W1> <R2>) -(hangs <W1> <R3>) (value <R1> <nl>) (value <R2> <n2>) - (value <W1> <nl> + <n2>) P2. Ropes over pulley. If a pulley system < P> has two ropes < RI > and < R2> over it, and the value (tension) associated with < RZ > is < nl > , then < nl > is also the value associated with rope < RZ > . P3. Rope hangs from or supports pulley. If there is a pulley system with ropes < RZ > and < R2> over it, and the pulley system hangs from a rope < R3 > , and c R1> and < R2 > have the values (tensions) < nl > and < n2 > associated with them, then the value (tension) associated with < R3 > is the sum of < nl > plus <n2>.

      Please explain to me how it is not evident to programmers that this is how we program.. we cannot hold more than seven items at a time. We cannot fracture. As Miller mentioned, 2 3 digit numbers are outside t capacity, but if you multiply them by paper w, if we free your memory, we can let the brain focus on the v

      If you assign the memory function to a diagram, you can let your brain concentrate on the manipulation function.

      Once we codify, we no longer have to keep the information in memory, for example : This hurts my brain

      There are five roads in Brown County. One runs from Abbeville to Brownsville by way of Clinton. One runs from Clinton to Derbyshire by way of Fremont. One runs from Fremont to Brownsville by way of Abbieville. That's all the roads in Brown County, and all the roads in and out of those towns.

      Which towns have roads connecting them directly to three other towns? Which towns have roads connecting them directly to only two other towns? How many towns must you pass through to get from Brownsville to Derbyshire?

      But if we diagram it to a map, all of this makes sense.

      "The learning of numbers and language must be subordinated ... Visual understanding is the essential and only valid means of teaching how to judge things correctly." ~ Johann Heinrich Pestalozzi

      Pestalozzi was the guy who designed the educational system in which Einstein, the most extraordinary visualization of his time, was born q

    1. to the bottom of the next image, about a fifth of a second later, like that. And they're getting faster and faster each time, and if I stack these guys up, then we see the differences; the increase in the speed is constant. And they say, "Oh, yeah. Constant acceleration. And how shall we

      For anyone interested in this I would also recommend anything regarding etoys https://www.youtube.com/watch?v=prIwpKL57dMhttp://www.squeakland.org/

  11. Jul 2023
      • Title
        • Is carbon tunnel vision real?
      • author Martin Daniel

      • This article introduces the concept of Product Environmental Footprint (PEF) as a way to validate if carbon tunnel vision is real

  12. Jun 2023
  13. May 2023
    1. It also includes some advice that might be obvious to professional programmers, but that might not be obvious to researchers or others who are just starting out
  14. Mar 2023
    1. él también viene al mundo equipadocon la capacidad de reconocer a sus congé-neres. La cara de la madre jugará un rolfundamental

      La visión se alimenta de luz y hace posible reconocer a nuestros congéneres y semejantes mediante el rostro y captar las innumerables formas producidas por la naturaleza y las creaciones culturales.

      La visión permite a los bebés interpretar los movimientos del rostro para crear significado, y los niños sordos la utilizan de una manera muy particular para aprender la lengua signada. (p. 32) lengua oral by Cabrejo

  15. Sep 2022
  16. Jun 2022
    1. park a crucial question that any leader who has to make a difficult, often unpopular, decision must ask themself: What must I do in this situation, given my position?
    2. Meaningful change, in contrast, supports a culture of integrity that is driven by formative institutional principles and that provides a structure for future growth and change.
    3. When we speak, we are afraid our words will not be heard or welcomed. But when we are silent, we are still afraid. So it is better to speak …”
  17. May 2022
    1. Building and sharing an app should be as easy as creating and sharing a video.

      This is where I think Glitch goes wrong. Why such a focus on apps (and esp. pushing the same practices and overcomplicated architecture as people on GitHub trying to emulate the trendiest devops shovelware)?

      "Web" is a red herring here. Make the Web more accessible for app creation, sure, but what about making it more accessible (and therefore simpler) for sharing simple stuff (like documents comprising the written word), too? Glitch doesn't do well at this at all. It feels less like a place for the uninitiated and more like a place for the cool kids who are already slinging/pushing Modern Best Practices hang out—not unlike societal elites who feign to tether themself to the mast of helping the downtrodden but really use the whole charade as machine for converting attention into prestige and personal wealth. Their prices, for example, reflect that. Where's the "give us, like 20 bucks a year and we'll give you better alternative to emailing Microsoft Office documents around (that isn't Google Sheets)" plan?

  18. Mar 2022
  19. Feb 2022
    1. Good example of developer tunnel vision and something akin to the consultant effect or Pournelle's Iron Law—the opposite of disintermediation.

    1. and if you want software that's any more niche than that

      That's the problem—thinking about this in terms of "wanting software". It's wanting to publish. Tech workers have an especially hard time understanding this.

      You're probably not under the impression that when the last person you heard of who got their book published finally pulled it off, they did it as a matter of wanting, say, an InDesign workflow versus something else. Because they weren't, and it didn't factor into their motivations at all—not even a little bit.

    1. The problem almost certainly starts with the conception of what we're doing as "building websites".

      When we do so, we mindset of working on systems

      If your systems work compromises the artifacts then it's not good work

      This is part of a broader phenomenon, which is that when computers are involved with absolutely anything people seem to lose their minds good sensibilities just go out the window

      low expectations from everyone everyone is so used to excusing bad work

      sui generis medium

      violates the principle of least power

      what we should be doing when grappling with the online publishing problem—which is what this is; that's all it is—is, instead of thinking in terms of working on systems, thinking about this stuff in such a way that we never lose sight of the basics; the thing that we aspire to do when we want to put together a website is to deal in

      documents and their issuing authority

      That is, a piece of content and its name (the name is a qualified name that we recognize as valid only when the publisher has the relevant authority for that name, determined by its prefix; URLs)

      that's it that's all a Web site is

      anything else is auxiliary

      really not a lot different from what goes on when you publish a book take a manuscript through final revisions for publication and then get an ISBN issued for it

      so the problem comes from the industry

      people "building websites" like politicians doing bad work and then their constituents not holding them accountable because that's not how politics works you don't get held accountable for doing bad work

      so the thing to do is to recognize that if we're thinking about "websites" from any other position things that technical people try to steer us in the direction of like selecting a particular system and then propping it up and how to interact with a given system to convince it to do the thing we want it to do— then we're doing it wrong

      we're creating content and then giving it a name

  20. Dec 2021
    1. Dreams or vision quests: among Iroquoian-speaking peoplesin the sixteenth and seventeenth centuries it was consideredextremely important literally to realize one’s dreams. ManyEuropean observers marvelled at how Indians would be willingto travel for days to bring back some object, trophy, crystal oreven an animal like a dog that they had dreamed of acquiring.Anyone who dreamed about a neighbour or relative’spossession (a kettle, ornament, mask and so on) couldnormally demand it; as a result, such objects would oftengradually travel some way from town to town. On the GreatPlains, decisions to travel long distances in search of rare orexotic items could form part of vision quests.34
      1. On ‘dream economies’ among the Iroquois see Graeber 2001: 145–9. David Graeber. 2001. Toward an Anthropological Theory of Value: The False Coin of Our Own Dreams. New York: Palgrave.

      These dreams and vision quests sound suspiciously familiar to Australian indigenous peoples' "dreaming" and could be incredibly similar to much larger and longer songlines in North American cultures.

  21. Nov 2021
    1. This is the cybercrud problem: advice and creation of systems, supposedlybased on technical requirements, whose categories and rigidities are unnecessary.In the worst cases they are not only unnecessary but wrong.

      Still a problem.

    2. After the conference dinner, Theodor Nelson of The Nelson Organization, Inc.,described a vision of what the computer's use in instruction might become, if onlywe could see beyond the "trivial horizons" of most computer peop:.e

      Still a problem.

  22. Oct 2021
    1. Cada aspecto de la visión, desde las proteínas de opsina hasta los ojos y las formas en que sirven al comportamiento animal, es increíblemente diverso. Solo con una perspectiva evolutiva se puede comprender y apreciar plenamente esta diversidad. En esta revisión, describo y explico la diversidad en cada nivel y trato de transmitir una comprensión de cómo el origen de la primera ops en hace unos 800 millones de años pudo iniciar la avalancha que produjo la asombrosa diversidad de ojos y visiones que vemos hoy. A pesar de la diversidad, muchos tipos de fotorreceptores, ojos y roles visuales han evolucionado varias veces de forma independiente en diferentes animales, revelando un patrón de evolución ocular estrictamente guiado por limitaciones funcionales e impulsado por la evolución de comportamientos gradualmente más exigentes.
    1. Este documento es interesante Debo agregarlo a la bibliografía de visión del color Este es un volumen especial dedicado el tema

  23. Sep 2021
    1. Around 1:45:50

      "Code is open source, but who cares? Nobody can understand it."


      Yeah, and it's useful. Like: people do stuff with it, but as soon as you run into problems with a library, it's just like... looking into the code is a mess... There's just a huge number of problems with it.

  24. Jun 2021
    1. Avoid 'global magic' or things that are defined outside of scope where they are not visible.

      From the commentary in the video "Workflow: Universal project folder structure"

      "I can intuit that this has something to do with[...]"

      "I look at this folder[...] and I get some sense[...]"

      "It's got this package dot bin thing, oh okay, so that means there's also a special command that I can run with this[...] you understand there is a command here"

  25. May 2021
    1. LifeInColor

      Se puede ver el color de las flores y la percepción de los insectos de esas flores, los insectos pueden ver UV

      Aparecen un abeja y una araña, la araña presenta cripsis del tipo coincidencia de fondo con la flor, el color de la araña de pende de pigmentos obtenidos por alimento.

  26. Apr 2021
    1. OpenCV Android Application Programming with OpenCV Mastering OpenCV with Practical Computer Vision Projects Practical OpenCV Learning OpenCV

      sách hay vọc OpenCV trên C++ và Java (Nhúng và di động)

  27. Mar 2021
    1. What I’d like more of is a social web that sits between these two extremes, something with a small town feel. So you can see people are around, and you can give directions and a friendly nod, but there’s no need to stop and chat, and it’s not in your face. It’s what I’ve talked about before as social peripheral vision (that post is about why it should be build into the OS).

      I love the idea of social peripheral vision online.

  28. Feb 2021
  29. Oct 2020
  30. Sep 2020
    1. The idea of a web browser being something we can comprehend, of a web page being something that more people can make, feels exciting to me.

      my personal hope is that we can build a more sensible coherent web, that exudes the machines inside of it, by better harkening towards custom elements ("webcomponents"). move the page from being a bunch of machines in javascript, to a bunch of machines in hypertext.

      and then build pages that start to expose & let the user play with the dom. start to build experiences that bridge the gap into the machine/page.

      and keep going. keep going. build wilder web experiences. build more machines. and keep building battlesuits for the user, out of more componenets, out of more web, to let them wrestle & tangle with & manipulate & experiment & hack on & see & observe & learn about the truthful, honest, direct hypertext that we all navigate.

  31. Jul 2020
    1. À l’inverse, ces artefacts, en privant la mémoire interne d’informations à synthétiser, ne risquent-ils pas de l’affaiblir et, à terme, de porter atteinte à notre façon de penser et à notre libre arbitre ?

      La question est centrale de l'article. L'auteur met l'accent sur l'importance de co-réflexion sur le theme avec cette question ce qui renforce qu'il s'agit plutôt d'une vision alarmiste.

    2. Réfléchir aux conséquences de l’externalisation de la mémoire humaine ne date pas d’hier.

      Retour à l'aspect historique. C'est plutôt un élément rhétorique pour souligner l'importance d'une réflexion commune sur la problématique. Ce qui nous renforce que c'est tout d'abord un article alarmiste.

  32. May 2020
  33. Apr 2020
    1. Installation in Windows Compatibility: > OpenCV 2.0 Author: Bernát Gábor You will learn how to setup OpenCV in your Windows Operating System!
    2. Here you can read tutorials about how to set up your computer to work with the OpenCV library. Additionally you can find very basic sample source code to introduce you to the world of the OpenCV. Installation in Linux Compatibility: > OpenCV 2.0
    1. OpenCV (Open Source Computer Vision Library) is an open source computer vision and machine learning software library. OpenCV was built to provide a common infrastructure for computer vision applications and to accelerate the use of machine perception in the commercial products. Being a BSD-licensed product, OpenCV makes it easy for businesses to utilize and modify the code. The library has more than 2500 optimized algorithms, which includes a comprehensive set of both classic and state-of-the-art computer vision and machine learning algorithms. These algorithms can be used to detect and recognize faces, identify objects, classify human actions in videos, track camera movements, track moving objects, extract 3D models of objects, produce 3D point clouds from stereo cameras, stitch images together to produce a high resolution image of an entire scene, find similar images from an image database, remove red eyes from images taken using flash, follow eye movements, recognize scenery and establish markers to overlay it with augmented reality, etc. OpenCV has more than 47 thousand people of user community and estimated number of downloads exceeding 18 million. The library is used extensively in companies, research groups and by governmental bodies. Along with well-established companies like Google, Yahoo, Microsoft, Intel, IBM, Sony, Honda, Toyota that employ the library, there are many startups such as Applied Minds, VideoSurf, and Zeitera, that make extensive use of OpenCV. OpenCV’s deployed uses span the range from stitching streetview images together, detecting intrusions in surveillance video in Israel, monitoring mine equipment in China, helping robots navigate and pick up objects at Willow Garage, detection of swimming pool drowning accidents in Europe, running interactive art in Spain and New York, checking runways for debris in Turkey, inspecting labels on products in factories around the world on to rapid face detection in Japan. It has C++, Python, Java and MATLAB interfaces and supports Windows, Linux, Android and Mac OS. OpenCV leans mostly towards real-time vision applications and takes advantage of MMX and SSE instructions when available. A full-featured CUDAand OpenCL interfaces are being actively developed right now. There are over 500 algorithms and about 10 times as many functions that compose or support those algorithms. OpenCV is written natively in C++ and has a templated interface that works seamlessly with STL containers.
  34. Oct 2019
    1. Franchise products are governed too much by economic logic and not enough by artistic vision

      Frankly this is because of the licensing issue as we know it. A company buys out a license to use Marvel characters in their films or games; in order to make the most money out of the license, maintain ownership of their license, and to avoid paying a ludicrous sum for it again, the company in question has to make media featuring marvel characters - hence the flow of sequels. Instead of artistic vision, address the licensing issue and the artistic vision will actually have room to flower.

  35. Apr 2019
  36. Mar 2019
    1. The main purpose of the Discovery IN is to provide interfaces and other user-facing services for data discovery across disciplines. We explore new and innovative ways of enabling discovery, including visualizations, recommender systems, semantics, content mining, annotation, and responsible metrics. We apply user involvement and participatory design to increase usability and usefulness of the solutions. We go beyond academia, involving users from all stakeholders of research data. We create FAIR and open infrastructures, following the FAIR principles complemented by the principles of open source, open data, and open content, thus enabling reuse of interfaces and user-facing services and continued innovation. Our main objectives are:
  37. Oct 2018
    1. malignant eye,

      Birds of prey have excellent eyesight. Scientists estimate that they can see two to eight times better than humans. They also have a third eyelid for protection.

    1. I saw the movement of content across media as an enhancement of the creative process. He saw it as a distraction or corruption.

      Points to a short-sightedness and tunnel vision in sections of media. Taking a focussed view on a very narrow area of a field, as opposed to a "world view" as advocated by the author.

    1. When irrationalism, as the counterplay of rationalism, talks about the things to which rationalism is blind, it does so only with a squint.

      Heidegger: "When irrationalism, as the counterplay of rationalism, talks about the things to which rationalism is blind, it does so only with a squint" || c.f. the narrowing / dimming of the optical field implied by the squint with the optics of disclosure and the rhetoric of blindness (de Man)

  38. Sep 2018
    1. Considero que este fragmento es importante ya que demuestra la visión de algunas de las empresas mexicanas, que es no abrirse a personas ajenas a la familia. Es una idea compartida en nuestro contexto y por lo que he checado en otros artículos también de los demás países de américa latina, es un dato interesante.

  39. Aug 2018
    1. By the year 2020 Estonian Smart City Cluster is internationally known as the leading developer and exporter of smart city solutions that are based on ICT and other technologies.
  40. Jun 2018
    1. Remember, the author made a more technical report on this topic. PDF here

    2. if we ever find the translation is dominant in a direction other than forward, we simply ignore that motion.

      Remember, this is just a heuristic

    3. Most Computer Vision algorithms are not complete without a few heuristics thrown in
    4. RANSAC. It is an iterative algorithm. At every iteration, it randomly samples five points from out set of correspondences, estimates the Essential Matrix, and then checks if the other points are inliers when using this essential matrix.
    5. T his step compensates for this lens distortion.
    6. For every pair of images, we need to find the rotation matrix RRR and the translation vector ttt, which describes the motion of the vehicle between the two frames.
    7. An efficient solution to the five-point relative pose problem
    1. our job is to construct a 6-DOF trajectory

      This is "forwards/backwards" in every major axis, and rotations in the same axis.

    1. Instead, we define the goal of the growth team as accelerating the realization of LinkedIn’s vision, which is to create economic opportunity for every member of the global workforce. Keeping this vision top of mind led us to set the right growth objectives and priorities.
    1. “Rap Genius is going be the fabric of the internet,” co-founder Mahbod Moghadam said in 2014 “Rap Genius is going be the fabric of the internet,” co-founder Mahbod Moghadam said in 2014. “We’re going to have annotations on other sites, so every other site in the world like the Wall Street Journal and the New York Times are going be Genius-powered and they’re going to have our annotations on them. And then the Genius platform will take over the internet; everyone’s most important statistic that they have in life is their Genius IQ.”
  41. Apr 2018
  42. Mar 2018
  43. Dec 2017
    1. Of another parcel of 153 acres near the former, and including a considerable eminence very favorable for the erection of a future observatory.

      This excerpt is very interesting as the reservation of land for the specific purpose of constructing an observatory seems to be very peculiar when considering the primary plans for the University. However, its inclusion is very relevant, as such reservations lead to the construction of the Leander McCormick Observatory, which currently sits on the summit of Mount Jefferson, commonly referred to as Observatory Hill. Though it took nearly 70 years for such plans to be carried out, the implementation of the McCormick Observatory has proved to be a prominent addition to the University, as it has helped enhance the education and has also served as a platform for astronomical research within the Astronomy Department.

      Link: http://astronomy.as.virginia.edu/research/observatories/mccormick

  44. Nov 2017
    1. As well might it be urged that the wild & uncultivated tree, hitherto yielding sour & bitter fruit only, can never be made to yield better: yet we know that the grafting art implants a new tree on the savage stock, producing what is most estimable both in kind & degree.

      I believe this metaphor to be very profound, as it challenges the statement within the previous sentences that, "man is fixed." The excerpt vividly depicts how a tree that yields sour and bitter fruit can be changed, through artistic processes, to produce a rather sweet fruit. These artistic processes serve to represent how man can be improved and changed through the role of education. Consequently, the Rockfish Gap Report discusses how education is crucial to the positive development of man, further illustrating Jefferson’s vision of the University of Virginia, where individuals will not only continue to grow as students, but also as cordial and honorable members of society. I believe that Jefferson’s vision has been fulfilled, as the University’s academic programs have enhanced the minds of those who have been, and are currently, students, while also providing them with a foundation of integrity and honor which will stay with them throughout their lives.

  45. Sep 2017
    1. Duncan (16) found that less than 5% of the figures in a typical textbook contain data. It is no wonder that students using a traditional and passive textbook do not know how to support their answers with data.

      Absolutely. The question is how to bridge the Perry Scheme so that students are more focused on evidence than the conclusion.

    2. However, if reading the book is a key component of class time and tests, students will use the text to help them construct their own understanding of the material.

      I agree. The text should be the organizing factor.

    3. Second, students must come to class prepared for classroom activities, which also facilitates studying for exams as the semester progresses.

      Hypothes.is also helps with this too

    4. First, student work outside of class needs to be intentionally and effectively structured.

      Hypothes.is is one tool that I use to achieve this goal.

    1. he admission of enlargement to any degree to which the institution may extend in future times.

      The Commissioners founded UVa with a vision and the confidence that the university will prosper and expand. They looked forward to the future and determine to make an impact--a core value still upheld by the university as of today.

  46. Aug 2017
  47. Jun 2017
    1. Who is Mistaken?Benjamin EysenbachMITbce@mit.eduCarl VondrickMITvondrick@mit.eduAntonio TorralbaMITtorralba@csail.mit.eduFigure 1: Can you determine who has a false belief about this scene? In this paper, we study how to recognize when a person in a short sequence is mistaken. Above, the woman is mistaken about the chair being pulled away from her.TimeFigure 1:Can you determine who believes something incorrectly in this scene?In this paper, we study how to recognizewhen a person in a scene is mistaken. Above, the woman is mistaken about the chair being pulled away from her in the thirdframe, causing her to fall down. Thered arrowindicates false belief. We introduce a new dataset of abstract scenes to studywhen people have false beliefs. We propose approaches to learn to recognizewhois mistaken andwhenthey are mistaken.AbstractRecognizing when people have false beliefs is crucial forunderstanding their actions. We introduce the novel prob-lem of identifying when people in abstract scenes have in-correct beliefs. We present a dataset of scenes, each visuallydepicting an 8-frame story in which a character has a mis-taken belief. We then create a representation of characters’beliefs for two tasks in human action understanding: pre-dicting who is mistaken, and when they are mistaken. Ex-periments suggest that our method for identifying mistakencharacters performs better on these tasks than simple base-lines. Diagnostics on our model suggest it learns importantcues for recognizing mistaken beliefs, such as gaze. We be-lieve models of people’s beliefs will have many

      Interesting

  48. Sep 2016
    1. The Tesla accident in May, researchers say, was not a failure of computer vision. But it underscored the limitations of the science in applications like driverless cars despite remarkable progress in recent years, fueled by digital data, computer firepower and software inspired by the human brain.

      Testing annotations. Interesting statement.

  49. Jul 2016
    1. Bird eyes have had eons longer to optimize. Along with their higher cone count, they achieve a far more regular spacing of the cells.
    1. Bird eyes have had eons longer to optimize. Along with their higher cone count, they achieve a far more regular spacing of the cells.
    1. Unsupervised Learning of 3D Structure from Images Authors: Danilo Jimenez Rezende, S. M. Ali Eslami, Shakir Mohamed, Peter Battaglia, Max Jaderberg, Nicolas Heess (Submitted on 3 Jul 2016) Abstract: A key goal of computer vision is to recover the underlying 3D structure from 2D observations of the world. In this paper we learn strong deep generative models of 3D structures, and recover these structures from 3D and 2D images via probabilistic inference. We demonstrate high-quality samples and report log-likelihoods on several datasets, including ShapeNet [2], and establish the first benchmarks in the literature. We also show how these models and their inference networks can be trained end-to-end from 2D images. This demonstrates for the first time the feasibility of learning to infer 3D representations of the world in a purely unsupervised manner.

      The 3D representation of a 2D image is ambiguous and multi-modal. We achieve such reasoning by learning a generative model of 3D structures, and recover this structure from 2D images via probabilistic inference.

    1. When building a unified vision system or gradually adding new capabilities to a system, the usual assumption is that training data for all tasks is always available. However, as the number of tasks grows, storing and retraining on such data becomes infeasible. A new problem arises where we add new capabilities to a Convolutional Neural Network (CNN), but the training data for its existing capabilities are unavailable. We propose our Learning without Forgetting method, which uses only new task data to train the network while preserving the original capabilities. Our method performs favorably compared to commonly used feature extraction and fine-tuning adaption techniques and performs similarly to multitask learning that uses original task data we assume unavailable. A more surprising observation is that Learning without Forgetting may be able to replace fine-tuning as standard practice for improved new task performance.

      Learning w/o Forgetting: distilled transfer learning

  50. Jun 2016
    1. Dynamic Filter Networks

      "... filters are generated dynamically conditioned on an input" Nice video frame prediction experiments.

    1. Atl=xtifl= 0MAXPOOL(RELU(CONV(Etl1)))l >0(1)^Atl=RELU(CONV(Rtl))(2)Etl= [RELU(Atl^Atl);RELU(^AtlAtl)](3)Rtl=CONVLSTM(Et1l;Rt1l;Rtl+1)(4)

      Very unique network structure. Prediction results look promising.

  51. May 2016
  52. Dec 2015
  53. May 2015
    1. il fait peur non pas parce qu’il commence à ressembler à l’humain, mais parce qu’au contraire, il se montre inflexible, aveugle au contexte, et, qu’il ait raison ou qu’il ait tort, capable de soumettre les hommes aux lois qu’ils ont eux-mêmes créés.
  54. Mar 2015
    1. ImageNet is an image database organized according to the WordNet hierarchy (currently only the nouns), in which each node of the hierarchy is depicted by hundreds and thousands of images. Currently we have an average of over five hundred images per node. We hope ImageNet will become a useful resource for researchers, educators, students and all of you who share our passion for pictures.
  55. Jan 2014
    1. An effective data management program would enable a user 20 years or longer in the future to discover , access , understand, and use particular data [ 3 ]. This primer summarizes the elements of a data management program that would satisfy this 20-year rule and are necessary to prevent data entropy .

      Who cares most about the 20-year rule? This is an ideal that appeals to some, but in practice even the most zealous adherents can't picture what this looks like in some concrete way-- except in the most traditional ways: physical paper journals in libraries are tangible examples of the 20-year rule.

      Until we have a digital equivalent for data I don't blame people looking for tenure or jobs for not caring about this ideal if we can't provide a clear picture of how to achieve this widely at an institutional level. For digital materials I think the picture people have in their minds is of tape backup. Maybe this is generational? New generations not exposed widely to cassette tapes, DVDs, and other physical media that "old people" remember, only then will it be possible to have a new ideal that people can see in their minds-eye.