Procedure learning refers to the task of identifying the key-steps and determining their logical order, given several videos of the same task. For both third-person and first-person (egocentric) videos, state-of-the-art (SOTA) methods aim at finding correspondences across videos in time to accomplish procedure learning. However, to establish temporal relationships within the sequences, these methods often rely on frame-to-frame mapping, or assume monotonic alignment of video pairs, leading to sub-optimal results. To this end, we propose to treat the video frames as samples from an unknown distribution, enabling us to frame their distance calculation as an optimal transport (OT) problem. Notably, the OT-based formulation allows us to relax the previously mentioned assumptions.

To this end, we propose to treat the video frames as samples from an unknown distribution, enabling us to frame their distance calculation as an optimal transport (OT) problem. Notably, the OT - based formulation allows us to relax the previously mentioned assumptions. To further improve performance, we enhance the OT formulation by introducing two regularization terms.

Procedure learning refers to the task of identifying the key-steps and determining their logical order, given several videos of the same task. For both third-person and first-person (egocentric) videos, state-of-the-art (SOTA) methods aim at finding correspondences across videos in time to accomplish procedure learning. However, to establish temporal relationships within the sequences, these methods often rely on frame-to-frame mapping, or assume monotonic alignment of video pairs, leading to sub-optimal results. To this end, we propose to treat the video frames as samples from an unknown distribution, enabling us to frame their distance calculation as an optimal transport (OT) problem. Notably, the OT-based formulation allows us to relax the previously mentioned assumptions.

Welcome to The Morning Shift, your roundup of the auto news you crave, all in one place every weekday morning. Here are the important stories you need to know. Everyone and their grandma is creating autonomous tech these days. Tesla, Audi, Cadillac, Mercedes and new players like Uber and Google are just a few of the companies developing software to make cars drive themselves eventually. But Toyota is hanging back.

The German transport ministry says General Motors' Opel division has been asked to appear before a commission looking into diesel emissions controls after an environmental group claimed two of its models are able to reduce pollution controls. The environmental group, DUH, says it has tested Opel's Zafira and Astra models and claims they reduce pollution controls at some speeds and temperatures. DUH wants them taken off the road. Opel says DUH's tests weren't objective or scientifically grounded, saying "our software was never designed to cheat or deceive." Apparently referring to the computer expert who examined the software for DUH, the company said that "the isolated conclusions of a hacker do not reflect the complex interdependencies of a modern exhaust after-treatment system."