English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Sequential matching using hand-crafted heuristics has been standard practicein route-based place recognition for enhancing pairwise similarity results fornearly a decade. However, precision-recall performance of these algorithmsdramatically degrades when searching on short temporal window (TW) lengths,while demanding high compute and storage costs on large robotic datasets forautonomous navigation research. Here, influenced by biological systems thatrobustly navigate spacetime scales even without vision, we develop a jointvisual and positional representation learning technique, via a sequentialprocess, and design a learning-based CNN+LSTM architecture, trainable viabackpropagation through time, for viewpoint- and appearance-invariant placerecognition. Our approach, Sequential Place Learning (SPL), is based on a CNNfunction that visually encodes an environment from a single traversal, thusreducing storage capacity, while an LSTM temporally fuses each visual embeddingwith corresponding positional data -- obtained from any source of motionestimation -- for direct sequential inference. Contrary to classical two-stagepipelines, e.g., match-then-temporally-filter, our network directly eliminatesfalse-positive rates while jointly learning sequence matching from a singlemonocular image sequence, even using short TWs. Hence, we demonstrate that ourmodel outperforms 15 classical methods while setting new state-of-the-artperformance standards on 4 challenging benchmark datasets, where one of themcan be considered solved with recall rates of 100% at 100% precision, correctlymatching all places under extreme sunlight-darkness changes. In addition, weshow that SPL can be up to 70x faster to deploy than classical methods on a 729km route comprising 35,768 consecutive frames. Extensive experimentsdemonstrate the... Baseline code available athttps://github.com/mchancan/deepseqslam

Sequential Place Learning: Heuristic-Free High-Performance Long-Term Place Recognition