funding
Robotic automation coming to plant nurseries
Eric Smalley writes in a Wired report that several companies are testing autonomous robots for plant nurseries that grow ornamental shrubs and trees. The robots use computer vision and other sensors to navigate and to pick up and deliver potted shrubs and trees. One startup, Harvest Automation, has raised $5 million in venture capital. A video of the robots in action is available here.
Related articles on Computer Vision Central:
- Computer vision for agriculture increases profits, reduces waste
- Google awards research grant in mobile crop surveillance
- Computer Vision used to target individual weeds
- Vision-based pruning system reduces labor costs in vineyards
- Senate-approved Department of Agriculture appropriations bill includes funding for computer vision staff
- Vision analysis can estimate crop yields earlier
U.S. funds research towards street intersection analyzer for the blind
The U.S. National Institutes of Health (NIH) is providing $400,000 in funding to Dr. James Coughlan, of the Smith-Kettlewell Eye Research Institute, to develop tools to help the visually impaired. Building on the team's existing computer vision software for detecting crosswalks and 'walk' lights, the new system will help the blind and visually impaired navigate across intersections via continous feedback. The goal is to deploy the system in the form of free smartphone apps for common platforms such as Android and iOS so that users will not have to carry a separate device.
Improving smartphone cameras
The U.S. National Science Foundation (NSF) has awarded $450,000 to Colorado-based company FiveFocal to improve the manufacturing of camera modules for mobile phones and other platforms. The grant is primarily targeted to increasing yields, which would reduce costs; however, it is likely that the improved manufacturing process monitoring will lead to higher quality as well. Increasing numbers of apps that require computer vision software are being written for smartphone platforms; higher and more consistent imaging quality will enhance the performance of many of those computer vision apps. A detailed description of the project is included below.
Combining knowledge with data for generalizable and robust visual learning
The U.S. National Science Foundation (NSF) has awarded a $200,000 grant to Rensselaer Polytechnic Institute (RPI) to study the combination of heterogeneous domain knowledge with data-driven machine learning algorithms for computer vision problems. The researchers, led by Dr. Qiang Ji, will systematically identify knowledge from different sources and develop methods of integrating it with data-driven algorithms.
Research abstract:
Computer vision has made tremendous progress in the past decades, partially enabled by the advanced machine learning techniques. But compared with human perception, computer vision remains primitive. One contributing factor for this is the data-driven nature of the current learning algorithms and their inability to incorporate any related knowledge. The data-driven methods tend to be database-specific and cannot generalize well to unseen data. This project addresses this issue through the introduction of a knowledge-augmented statistical learning framework. Within this framework, knowledge and data can be systematically exploited, captured, and are principally integrated to jointly train a vision algorithm. Developing such a framework, however, is challenging since the domain knowledge often exists in different and diverse formats, typically inaccessible to the data-driven statistical machine learning methods. To overcome this challenge, the research team systematically converts domain knowledge into either the constraints on the model or into pseudo-data, whereby they can be incorporated into the statistical learning methods. The project includes systematic identification of knowledge from different sources and concrete mechanisms to capture the knowledge and to convert them into formats easily accessible to the automatic machine learning methods. The project also involves demonstrating the effectiveness of the proposed framework for certain computer vision problems.
Debugging machine visual recognition via humans in the loop
The U.S. National Science Foundation (NSF) has awarded a $150,000 grant to Devi Parikh of the Toyota Technological Institute at Chicago to investigate the use of humans to replace parts of a computer vision object recognition pipeline. The goal of this "human debugging" experiment is to identify "weak links in computational models via humans in the loop."
Abstract of the research:
The problem of visual recognition is fundamental towards the goal of automatic image understanding. While a large number of efforts have been made in the computer vision community, machine performance at these tasks remains significantly inferior to human ability.
The overarching goal of this project is to leverage the best known visual recognition system - the human visual recognition system. This project employs a "Human Debugging" paradigm to replace various components of a machine vision pipeline with human subjects, and examines the resultant effect on recognition performance. Meaningful comparisons provide valuable insights and pinpoint aspects of the machine vision pipeline that are performance bottlenecks and require future research efforts. Specifically, the project considers the problems of image classification and object detection, and explores the roles of local and global information, as well part-detection, spatial modeling and contextual reasoning (including non-maximal suppression) for these problems respectively.
U.S. funding research on 'shoulder surfing' attack mitigation
The U.S. National Science Foundation (NSF) has awarded $150,000 to Jan-Michael Frahm and Fabian Monrose of the University of North Carolina at Chapel Hill to investigate computer security attacks based on viewing mobile device screens and keyboards. (See "Thinkst warns of computer vision shoulder surfing attack on tablets" for an example of this kind of capability.) The funded research will characterize vulnerabilities and propose mitigation mechanisms.
Title: Automatic Reconstruction of Typed Input from Compromising Reflections
Abstract:
This project explores computer vision techniques aimed at exploiting compromising reflections associated with data input in mobile electronic devices such as smart phones. The ubiquity of these personal communication devices and their growing roles in data manipulation tasks, make unintended visual emanations an exploitable liability to data security. Nevertheless, there is still a gap in understanding of both the limitations of these techniques as well as the availability of effective mitigation mechanisms. It is the goal of this work to contribute to filling this conceptual gap.
The study builds upon recent state of the art techniques for automatic reconstruction of typed input from compromising reflections, comprising of robust keystroke event detection and classification mechanisms coupled to natural language processing modules. Such paradigm is both effective and amenable to low cost implementation in commodity devices. Based on these new developments, threat scenarios are no longer restricted to controlled scenarios using specialized equipment, but rather consist of highly flexible and possibly impromptu attacks. The project develops advanced cross-platform data input transcription prototypes used within a threat validation framework. This framework provides a characterization of both threat scenario operational limitations (e.g., imaging resolution, scene illumination, computational requirements) as well as the performance characteristics (e.g., robustness, accuracy) of the different vulnerability exploitation mechanisms. Moreover, the results of the analysis of diverse threat scenarios are being used to identify and develop appropriate mitigation mechanisms when possible.
Yale University awarded grant for a neural theory of 3D shape perception
The U.S. National Science Foundation is awarding a two-year, $300,000 grant to Professor Steven Zucker of Yale University to investigate a neural theory of 3D shape perception. The grant is part of a U.S.-German collaboration. The abstract of the research is included below.
Abstract
How the brain estimates the 3D shape of objects in our surroundings remains one of the most significant challenges in visual neuroscience. The information provided by the retina is fundamentally ambiguous, because many different combinations of 3D shape, illumination and surface reflectance are consistent with any given image. Despite this ambiguity, the visual system is extremely adept at estimating 3D shape across a wide range of viewing conditions, something that no extant machine vision system can do. The long-term goal of the project is to develop a computational model in neural terms to explain how 3D shape is estimated in the primate visual system. It will build upon the responses of cells early in visual cortex (V1) and develop models of how they can be organized into mid-level configurations that specify 3D shape properties. Importantly, the project will also measure human perception of 3D shape in a series of psychophysical experiments designed to test specific predictions, bringing together the complementary expertise of Roland W. Fleming (Giessen University: human perception, psychophysics) and Steven W. Zucker (Yale University: computational vision, computational neuroscience). The results should provide a deeper understanding of visual circuit properties in the ventral processing stream; they should provide models for 3D computer vision and graphics; and they may pave the way for the development of rehabilitation strategies for patients with visual deficits.
TextureCam: Onboard Image Analysis for Autonomous Astrobiology
The U.S. space agency, NASA, is funding the development of TextureCam, a space exploration camera system with texture analysis algorithms embedded in integrated hardware. It is intended to increase the analyses that can be done onsite, thus enhancing the ability of space exploration vehicles such as Mars rovers to autonomously select targets of study. That is important due to the long communication time lags and communication blackouts that extraplanetary space missions are subject to.
TextureCam: Onboard Image Analysis for Autonomous Astrobiology Survey
Principal Investigator: David Thompson – Jet Propulsion Laboratory
TextureCam is a visible-wavelength imager with integrated texture analysis hardware. It is a new class of imaging instrument with texture channels that differentiate and map habitat-relevant surfaces. Here the term “texture” is used in the computer vision sense to signify statistical patterns of image pixels. These numerical signatures can automatically distinguish geologically relevant elements such as roughness, pavement coatings, regolith characteristics, sedimentary fabrics, and differential weathering in outcrop. This onboard ‘data understanding’ capability minimizes problems caused by communications delays, blackouts, and narrow bandwidth data transfer, and can benefit science return by summarizing features encountered during travel and directing autonomous instrument deployment for targets of opportunity.
NSF funds research on filling occlusions in point clouds
ClearEdge3D announced that is has been awarded a grant by the U.S. National Science Foundation (NSF) to investigate methods for improving the generation of 3D point clouds. According to the company, "due to the 'line of sight' limitations in laser scanning technology, there are always occluded regions of point cloud data that must be manually modeled. The core goal of the research grant is to devise algorithms that will automatically populate these occluded regions based on the feature pattern in the surrounding area."
Startup Magisto initiates automatic video editing service, receives $5.5 million
Computer vision startup Magisto is now making its automated video editing service publicly available. Previously available only to invitation-only members of a beta-test group, the Magisto algorithms take unedited home videos, search for the most interesting parts, and edit them into a movie. According to cnet news, "the system is designed to recognize faces -- even the most familiar faces -- as well as landscapes, animals, and movements such as zoom shots or action sequences." Magisto also received $5.5 in series B funding led by Li Ka-shing's Horizons Ventures.