BLOG

A new approach to image search uses images returned by traditional search methods as nodes in a graph neural network through which similarity signals are propagated, achieving improved ranking in cross-modal retrieval.

A new greedy, brute-force solution improves fairness in ranking and encompasses realistic scenarios with multiple, unknown protected groups.

By incorporating a set of reusable constructs that enable advanced functionality, Flow makes platform-agnostic application modelling, creation, distribution and maintenance easy.

How mid-size sets of captioned images can rival with large-scale labelled image sets to learn generic representations

A novel efficient model for whole-body 3D pose estimation (including bodies, hands and faces), that is trained by mimicking the output of hand-, body- and face-pose experts.

A new method called FORCE achieves extreme sparsity in artificial neural networks by progressively removing up to 99.9% of parameters at initialization, making it a promising candidate for training networks on edge devices (like drones or smartphones).

Announcing 'kapture' the open source release of a new SfM and visual localization data format. Available for ECCV2020 workshop challenge on visual localization.

ACL 2020 was held virtually the week of July 6th. As a combined industry/academic research team, we grouped information and papers on the topics we considered emerging at this year's conference.

A overview of current research methods in lifelong learning (rehearse, grow, regularize), the advantages and drawback of each and where we see future work heading.

With Ambient Wanderer, we aim to inform the design of systems that can meet the requirements of on-the-go users looking to discover relevant points of interest, and to demonstrate the benefits of using a hybrid Wizard of Oz approach to investigate POI recommender systems.

A first-of-its-kind architecture that, based on a single image, predicts how a robot can pick up objects from within any scene could revolutionize applications in AR/VR and robotics.

1^st in autonomous vehicle challenge, 2^nd in local feature and 4^th in handheld devices R2D2 and KAPTURE demonstrate generalizability and wide variety of applications