Are We There Yet? Challenging SeqSLAM on a 3000 km Journey Across All Four Seasons

By Niko Sünderhauf, Peer Neubert, Peter Protzel. 2013

331 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/88a00eb1eb6b1a7df54ed0ac668afde26122af1b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

It's not easy seeing green: Lighting-resistant stereo Visual Teach & Repeat using color-constant images (2015, unknown): “Naseer et. al [13] compare sequences of images by building a directed acyclic graph of possible matches over the entire trajectory and a graph search to find the best match.”
Robust visual SLAM across seasons (2015, methodology): “We use the open source implementations of FABMAP [10] and SeqSLAM [23].”
Visual Place Recognition: A Survey (2016, background): “However, as for other description methods, too drastic a change in appearance will cause system failure [111] and whole-image descriptors also suffer from the additional problem of sensitivity to viewpoint change [198].”
Illumination Invariant Imaging : Applications in Robust Vision-based Localisation , Mapping and Classification for Autonomous Vehicles (2014, background): “However, localisation performance is limited to topological place recognition, and is critically dependent on viewpoint direction when revisiting locations [18], making sequence-based approaches unsuitable as the sole localisation source for autonomous road vehicles.”
Fusion and binarization of CNN features for robust topological localization across seasons (2016, methodology): “The experiments for SeqSLAM are performed with OpenSeqSLAM [24].”
Place Recognition with ConvNet Landmarks: Viewpoint-Robust, Condition-Robust, Training-Free (2015, background): “Place Recognition: Methods that address the place recognition problem span from matching sequences of images [27, 17, 40, 33, 29], transforming images to become invariant against common scene changes such as shadows [6, 43, 25, 24, 21], learning how environments change over time and predicting these”
Benchmarking 6DOF Outdoor Visual Localization in Changing Conditions (2018, methodology): “In order to measure the benefit of using multiple images for pose estimation, we evaluate two approaches: OpenSeqSLAM [65] is based on image retrieval and enforces that the images in the sequence are matched in correct order.”
On the performance of ConvNet features for place recognition (2015, methodology): “We demonstrate the gain in performance by running OpenSeqSLAM [17] on the distance matrix created by the cosine distances on conv3 and show the results in Fig.”