(Press-News.org) PROVIDENCE, R.I. [Brown University] -- Researchers from Brown and Johns Hopkins universities have come up with a new way to evaluate how well computers can divine information from images. The team describes its new system as a "visual Turing test," after the legendary computer scientist Alan Turing's test of the extent to which computers display human-like intelligence.
"There have been some impressive advances in computer vision in recent years," said Stuart Geman, the James Manning Professor of Applied Mathematics at Brown. "We felt that it might be time to raise the bar in terms of how these systems are evaluated and benchmarked."
Traditional computer vision benchmarks tend to measure an algorithm's performance in detecting objects within an image (the image has a tree, or a car or a person), or how well a system identifies an image's global attributes (scene is outdoors or in the nighttime).
"We think it's time to think about how to do something deeper -- something more at the level of human understanding of an image," Geman said.
For example, it's one thing to be able to recognize that an image contains two people. But to be able to recognize that the image depicts two people walking together and having a conversation is a much deeper understanding. Similarly, describing an image as depicting a person entering a building is a richer understanding than saying it contains a person and a building. The system Geman and his colleagues developed, described this week in the Proceedings of the National Academy of Sciences, is designed to test for such a contextual understanding of photos.
It works by generating a string of yes or no questions about an image, which are posed sequentially to the system being tested. Each question is progressively more in-depth and based on the responses to the questions that have come before.
For example, an initial question might ask a computer if there's a person in a given region of a photo. If the computer says yes, then the test might ask if there's anything else in that region -- perhaps another person. If there are two people, the test might ask: "Are person1 and person2 talking?"
As a group, the questions are geared toward gauging the computer's understanding of the contextual "storyline" of the photo. "You can build this notion of a storyline about an image by the order in which the questions are explored," Geman said.
Because the questions are computer-generated, the system is more objective than having a human simply query a computer about an image. There is a role for a human operator, however. The human's role is to tell the test system when a question is unanswerable because of the ambiguities of the photo. For instance, asking the computer if a person in a photo is carrying something is unanswerable if most of the person's body is hidden by another object. The human operator would flag that question as ambiguous.
The first version of the test was generated based on a set of photos depicting urban street scenes. But the concept could conceivably be expanded to all kinds of photos, the researchers say.
Geman and his colleagues hope that this new test might spur computer vision researchers to explore new ways of teaching computers how to look at images. Most current computer vision algorithms are taught how to look at images using training sets in which objects are annotated by humans. By looking at millions of annotated images, the algorithms eventually learn how to identify objects. But it would be very difficult to develop a training set with all the possible contextual attributes of a photo annotated. So true context understanding may require a new machine learning technique.
"As researchers, we tend to 'teach to the test,'" Geman said. "If there are certain contests that everybody's entering and those are the measures of success, then that's what we focus on. So it might be wise to change the test, to put it just out of reach of current vision systems."
INFORMATION:
La Jolla, Calif., March 9, 2015 - Researchers at Sanford-Burnham Medical Research Institute (Sanford-Burnham) have discovered a mechanism that explains why some breast cancer tumors respond to specific chemotherapies and others do not. The findings highlight the level of glutamine, an essential nutrient for cancer development, as a determinant of breast cancer response to select anticancer therapies, and identify a marker associated with glutamine uptake, for potential prognosis and stratification of breast cancer therapy.
"Our study indicates that a protein called RNF5 ...
ANN ARBOR, Mich. -- The anti-inflammatory class of drugs NSAIDs have shown great promise in preventing cancers including colon, esophagus and skin. However, they can increase the risks of heart attacks, ulcers and rare but potentially life-threatening bleeds.
A new study suggests there may be ways to reduce these dangerous side effects.
Collaborators from the University of Michigan, the National Cancer Institute and the University of Alabama looked at naproxen, which is known to have a lower cardiovascular risk than other NSAIDs. Naproxen, like most NSAIDs and aspirin, ...
Parvoviruses are a class of viruses that normally infect rodents; in humans, they do not cause any disease symptoms. However, they are able to infect and kill cancer cells. The details behind this biological selectivity on the part of the viruses have not been understood until now. "Since the viruses might soon play a role in cancer medicine, it is important to know why they replicate exclusively in tumor cells in humans," says virologist Dr. Jürg Nüesch from the German Cancer Research Center (Deutsches Krebsforschungszentrum, DKFZ).
In order to complete their ...
Was there a single migration event or gradual mixing of cultures that gave rise to modern Japanese?
According to current theory, about 2,000-3,000 years ago, two populations, the hunter-gatherer Jomon from the Japanese archipelago, and the agricultural Yayoi from continental East Asia, intermingled to give rise to the modern Japanese population. However, some researchers have suggested otherwise, with the Jomon culture gradually transformed into the Yayoi culture without large migrations into modern day Japan.
To resolve the controversy, researchers Oota, Mano, Nakagome ...
The generation of neurons (neurogenesis) in humans is predominantly limited to development; in the adult stage it takes place in only a few regions of the brain. These regions contain neural stem cells that generate neurons in a process with various intermediary stages.
Stem cell renewal is limited - total number drops
Until now it was thought that maintaining the stem cell pool was based on the self-renewal of individual stem cells. The team of scientists headed by Dr. Jovica Ninkovic and Professor Dr. Magdalena Götz were able to refute this: Both the self-renewal ...
There is a rapidly growing momentum driving the development of mobile payment systems. New research has shown systems, such as the Bristol Pound, can have a positive effect on the local community by encouraging consumers to support and value their local businesses.
The paper, which will be presented at this month's ACM Conference on Computer-Supported Cooperative Work and Social Computing (ACM CSCW 2015) [14-18 March], looks at user interactions and practices around mobile payment systems.
The research, led by Dr Mark Perry at Brunel University London and visiting ...
Tourists are familiar with the Lodoicea maldivica palm, also called coco de mer, mainly because of their bizarrely shaped fruits. Scientists, however, are fascinated by the huge plants - which are abundant on the Seychelles islands of Praslin and Curieuse - for entirely different reasons.
The coco de mer palm engages in a lot of effort for reproduction, producing large amounts of pollen and huge fruits that cannot be spread around, but rather fall to the ground at the base. "This is nan enormous commitment of energy in very nutrient-poor soil - it does not really make ...
The central dogma of molecular biology describes the flow of genetic information. It was first described by Francis Crick in 1956 as one-way traffic: as: "DNA makes RNA and RNA makes protein."
A recent paper published in Mutation Research - Fundamental and Molecular Mechanisms of Mutagenesis, however suggests that, rather than being a one-way street, DNA-directed RNA transcription may have profound adaptability. The authors of the paper showed a conceptually novel relationship between the genotype (DNA) and the phenotype (the products of the transcription of DNA).
The ...
This news release is available in German. A momentary lapse of concentration is all it takes for a finger to become trapped or sprain an ankle - and it hurts. Pain is the body's protective mechanism and a complex neurological phenomenon. Moreover, ongoing pain in the sense of chronic pain can be a disease. Scientists from Technische Universität München (TUM) have now demonstrated that already during a few minutes of ongoing pain, the underlying brain activity changes by shifting from sensory to emotional processes.
In their experiments, Prof. Markus Ploner, ...
Until now there have been few molecular epidemiological studies regarding the effects of weight changes on metabolism in the general population. In a recent study conducted and funded within the framework of the Competence Network Obesity, researchers at the Institute of Epidemiology II at Helmholtz Zentrum München (HMGU) evaluated molecular data of the KORA study*. "Techniques such as metabolomics and transcriptomics allow the simultaneous determination of a variety of low molecular weight metabolites or gene activities (transcripts of genes) using high-throughput ...
