Tracking objects of interest in a video is one of the most popular and widely applicable problems in computer vision.

The video object segmentation (VOS) task involves the segmentation of an object over time based on a single initial mask. Current state-of-the-art approaches use a memory of previously processed frames and rely on matching to estimate segmentation masks of subsequent frames. Lacking any adaptation mechanism, such methods are prone to test-time distribution shifts. This work focuses on matching-based VOS under distribution shifts such as video corruptions, stylization, and sim-to-real transfer. We explore test-time training strategies that are agnostic to the specific task as well as strategies that are designed specifically for VOS.

The video object segmentation (VOS) task involves the segmentation of an object over time based on a single initial mask. Current state-of-the-art approaches use a memory of previously processed frames and rely on matching to estimate segmentation masks of subsequent frames. Lacking any adaptation mechanism, such methods are prone to test-time distribution shifts. This work focuses on matching-based VOS under distribution shifts such as video corruptions, stylization, and sim-to-real transfer. We explore test-time training strategies that are agnostic to the specific task as well as strategies that are designed specifically for VOS.

The general domain of video segmentation is currently fragmented into different tasks spanning multiple benchmarks. Despite rapid progress in the state-of-the-art, current methods are overwhelmingly task-specific and cannot conceptually generalize to other tasks. Inspired by recent approaches with multi-task capability, we propose TarViS: a novel, unified network architecture that can be applied to any task that requires segmenting a set of arbitrarily defined 'targets' in video. Our approach is flexible with respect to how tasks define these targets, since it models the latter as abstract 'queries' which are then used to predict pixel-precise target masks. A single TarViS model can be trained jointly on a collection of datasets spanning different tasks, and can hot-swap between tasks during inference without any task-specific retraining. To demonstrate its effectiveness, we apply TarViS to four different tasks, namely Video Instance Segmentation (VIS), Video Panoptic Segmentation (VPS), Video Object Segmentation (VOS) and Point Exemplar-guided Tracking (PET). Our unified, jointly trained model achieves state-of-the-art performance on 5/7 benchmarks spanning these four tasks, and competitive performance on the remaining two. Code and model weights are available at: https://github.com/Ali2500/TarViS

Out-of-distribution (OOD) detection has received much attention lately due to its importance in the safe deployment of neural networks. One of the key challenges is that models lack supervision signals from unknown data, and as a result, can produce overconfident predictions on OOD data. Previous approaches rely on real outlier datasets for model regularization, which can be costly and sometimes infeasible to obtain in practice. In this paper, we present VOS, a novel framework for OOD detection by adaptively synthesizing virtual outliers that can meaningfully regularize the model's decision boundary during training. Specifically, VOS samples virtual outliers from the low-likelihood region of the class-conditional distribution estimated in the feature space. Alongside, we introduce a novel unknown-aware training objective, which contrastively shapes the uncertainty space between the ID data and synthesized outlier data. VOS achieves state-of-the-art performance on both object detection and image classification models, reducing the FPR95 by up to 7.87% compared to the previous best method. Code is available at https://github.com/deeplearning-wisc/vos.

Enterprise Times spoke to Bas de Vos, Director of IFS Labs. He has been in charge of IFS Labs at arguably one of the most exciting times for technology research and development. The conversation was animated and full of information about what the Labs has achieved, what it is working on now and where de Vos sees technology potentially go in the future. We discussed Artificial Intelligence and de Vos views AI as a set of technologies that enable people to make products better. As with most things, he reflects the pragmatic view that IFS has around both R&D and new technology.

TROY, MI – The future of mobility will be data driven, says Glen De Vos, chief technical officer of Aptiv, and the former Delphi electronics unit wants to be the engine behind the transformation. "It is a big change for us," De Vos says of Aptiv's new role in the industry as chiefly a software company after decades of supplying automakers with components and parts hardware now considered low-margin commodities. De Vos would be happy to keep selling the hardware, too, and the Gillingham, U.K.-based company likely will for many years to come, but it will take a back seat to the software its customers need to raise the connectivity, improve the safety and reduce the unfavorable environmental impact of the vehicles they make. The supplier officially launched Dec. 5 with Kevin Clark, president and CEO of Aptiv, saying the company has an opportunity to play a large role in an industry transformation underpinned by the promise of autonomous vehicles. "Aptiv is built on a strong foundation of industry firsts and has the knowledge, capability and agility to win with traditional OEM customers and emerging mobility players," Clark says of the former Delphi Automotive.

With the growth of artificial intelligence, the fears of many have started to become a reality as jobs start falling to the machines. Many technologists and government officials will see the emergence of the digital era as a chance to score political points through job creation, however one has to ask whether the scale of the challenge is appreciated. For Dik Vos, CEO at SQS, the rapid evolution to the connected economy could see as many as 30% of the population being made redundant and in need to retraining. "The real impact will be seen in five to seven years but already in 2017 there will be greater disruption to the community and more jobs taken over by software," said Vos in an interview with Telecoms.com. "There will be a huge group of people who will be caught out when the rate of development goes much faster than we anticipated."

Auto parts and electronics maker Delphi Corp. has signed a deal with Intel to buy high-powered computer processors for Delphi's future autonomous vehicle systems. Delphi says Intel's added computing capacity will be needed as autonomous car systems gather and store more and more information while expanding their ability to deal with situations on real roads. In August, Delphi announced that it had joined with Israeli software maker Mobileye to develop the building blocks for a fully autonomous car in about two years. Intel Corp. will supply Delphi with high-capacity computers needed to process input from radar, cameras and laser sensors as well as maps of roadside landmarks. Glen De Vos, vice president of Delphi's business-services unit, says the Intel deal gives the company everything it needs to develop an autonomous-driving package to sell to automakers. Delphi makes its own radar and laser sensors and uses Mobileye's cameras and software.

The latest Google drones have just started taking flight in the real world. But the team behind the technology is slowing down, trimming headcount and shelving initiatives as the experimental unit becomes the latest target of tightening budgets across parent company Alphabet Inc. Project Wing, a unit of Alphabet's X research lab, nixed a partnership with coffee giant Starbucks Corp., according to people familiar with the decision. Following the departure of project leader Dave Vos in October, the unit also froze hiring and began asking some staff to seek jobs elsewhere in the company, according to some of those people. They asked not to be identified speaking about private company moves. The decisions are part of a broader Alphabet effort to rein in spending and try to turn more experimental projects from loss-making risky bets into real businesses.