Retrieval-based Localization Based on Domain-invariant Feature Learning under Changing Environments

@article{Hu2019RetrievalbasedLB,
  title={Retrieval-based Localization Based on Domain-invariant Feature Learning under Changing Environments},
  author={Hanjiang Hu and Hesheng Wang and Zhe Liu and Chenguang Yang and Weidong Chen and Le Xie},
  journal={2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  year={2019},
  pages={3684-3689}
}
  • Hanjiang Hu, Hesheng Wang, Le Xie
  • Published 23 September 2019
  • Computer Science
  • 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
Visual localization is a crucial problem in mobile robotics and autonomous driving. One solution is to retrieve images with known pose from a database for the localization of query images. However, in environments with drastically varying conditions (e.g. illumination changes, seasons, occlusion, dynamic objects), retrieval-based localization is severely hampered and becomes a challenging problem. In this paper, a novel domain-invariant feature learning method (DIFL) is proposed based on… 

Figures and Tables from this paper

Domain-Invariant Similarity Activation Map Contrastive Learning for Retrieval-Based Long-Term Visual Localization
TLDR
A novel gradient-weighted similarity activation mapping loss (Grad-SAM) is incorporated for finer localization with high accuracy and a new adaptive triplet loss is proposed to boost the contrastive learning of the embedding in a self-supervised manner.
DASGIL: Domain Adaptation for Semantic and Geometric-Aware Image-Based Localization
TLDR
This paper proposes a novel multi-task architecture to fuse the geometric and semantic information into the multi-scale latent embedding representation for visual place recognition and outperforms state-of-the-art baselines for retrieval-based localization and large-scale place recognition under the challenging environment.
Domain-invariant NBV Planner for Active Cross-domain Self-localization
TLDR
A novel multi-encoder deep convolutional neural network enables to detect domain invariant pole-like landmarks and is used as the sole input to a model-free deep reinforcement learning -based domain-invariant NBV planner that attains consistent performance over different domains without requiring the expensive task of training data collection and retraining.
A Survey on Deep Visual Place Recognition
TLDR
A survey of the state-of-the-art of research on visual place recognition, focusing on how it has been shaped by the recent advances in deep learning, is presented.
Dark Reciprocal-Rank: Boosting Graph-Convolutional Self-Localization Network via Teacher-to-student Knowledge Transfer
TLDR
A novel teacher-to-student knowledge-transfer scheme based on rank matching is introduced, in which the reciprocal-rank vector output by an off-the-shelf state-of- the-art teacher self-localization model is used as the dark knowledge to transfer.
Fault-Diagnosing Monocular-SLAM for Scale-Aware Change Detection
TLDR
Experiments on challenging cross-season ICD using publicly available NCLT dataset, and comparison against state-of-the-art ICD algorithms validate the efficacy of the proposed FD approach with/without combining AD and/or PC.
Learning a Robust Hybrid Descriptor for Robot Visual Localization
TLDR
The novel hybrid descriptor for long-term visual localization is proposed, which is generated by combining a semantic image descriptor extracted from segmentation images and an image descriptors extracted from RGB images with a certain weight, and then trained by a convolutional neural network.
Explicit feature disentanglement for visual place recognition across appearance changes
TLDR
This work uses an adversarial network with two discriminators to disentangle domain-related features and domain- unrelated features from images, and the domain-unrelated features are used as descriptors in place recognition.
Boosting Self-localization with Graph Convolutional Neural Networks
TLDR
A novel knowledge transfer framework is developed that introduces an arbitrary self-localization model as the teacher to train the GCN-based self-Localization system i.e., the student and results reveal that the proposed method outperforms existing comparative methods and teacher self- localization systems.
...
...

References

SHOWING 1-10 OF 24 REFERENCES
Night-to-Day Image Translation for Retrieval-based Localization
TLDR
This paper proposes ToDayGAN – a modified image-translation model to alter nighttime driving images to a more useful daytime representation, and improves localization performance by over 250% compared the current state-of-the-art, in the context of standard metrics in multiple categories.
A Multi-Domain Feature Learning Method for Visual Place Recognition
TLDR
This work proposes an end-to-end conditional visual place recognition method that introduces the multi-domain feature learning method (MDFL) to capture multiple attribute-descriptions for a given place, and uses a feature detaching module to separate the environmental condition-related features from those that are not.
NetVLAD: CNN Architecture for Weakly Supervised Place Recognition
TLDR
A convolutional neural network architecture that is trainable in an end-to-end manner directly for the place recognition task, and significantly outperforms non-learnt image representations and off-the-shelf CNN descriptors on two challenging place recognition benchmarks.
Multimodal Unsupervised Image-to-Image Translation
TLDR
A Multimodal Unsupervised Image-to-image Translation (MUNIT) framework that assumes that the image representation can be decomposed into a content code that is domain-invariant, and a style code that captures domain-specific properties.
Benchmarking 6DOF Outdoor Visual Localization in Changing Conditions
TLDR
This paper introduces the first benchmark datasets specifically designed for analyzing the impact of day-night changes, weather and seasonal variations, as well as sequence-based localization approaches and the need for better local features on visual localization.
Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks
TLDR
This work presents an approach for learning to translate an image from a source domain X to a target domain Y in the absence of paired examples, and introduces a cycle consistency loss to push F(G(X)) ≈ X (and vice versa).
Unsupervised Image-to-Image Translation Networks
TLDR
This work makes a shared-latent space assumption and proposes an unsupervised image-to-image translation framework based on Coupled GANs that achieves state-of-the-art performance on benchmark datasets.
FAB-MAP: Probabilistic Localization and Mapping in the Space of Appearance
TLDR
A probabilistic approach to the problem of recognizing places based on their appearance that can determine that a new observation comes from a previously unseen place, and so augment its map, and is particularly suitable for online loop closure detection in mobile robotics.
Aggregating local descriptors into a compact image representation
TLDR
This work proposes a simple yet efficient way of aggregating local image descriptors into a vector of limited dimension, which can be viewed as a simplification of the Fisher kernel representation, and shows how to jointly optimize the dimension reduction and the indexing algorithm.
The CMU Visual Localization Data Set
  • http://3dvis.ri.cmu.edu/data-sets/localization, 2011.
  • 2011
...
...