Spatial-Aware Feature Aggregation for Image based Cross-View Geo-Localization

Authors: Yujiao Shi, Liu Liu, Xin Yu, Hongdong Li

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on standard benchmark datasets show significant performance boosting, achieving more than doubled recall rate compared with the previous state of the art.
Researcher Affiliation Academia Yujiao Shi, Liu Liu, Xin Yu, Hongdong Li Australian National University, Canberra, Australia. Australian Centre for Robotic Vision, Australia. {firstname.lastname}@anu.edu.au
Pseudocode No The paper describes its methods in narrative text and uses diagrams (Figure 2, Figure 3) but does not contain any formal pseudocode or algorithm blocks.
Open Source Code Yes 1The code of this paper is available at https://github.com/shiyujiao/SAFA.
Open Datasets Yes Our experiments are conducted on two standard benchmark datasets: CVUSA [24] and CVACT [11], where ground images are panoramas.
Dataset Splits Yes CVUSA and CVACT are both cross-view datasets, and each dataset contains 35, 532 ground-and-aerial image pairs for training. CVUSA provides 8, 884 image pairs for testing and CVACT provides the same number of pairs for validation (denoted as CVACT_val). Besides, CVACT also provides 92, 802 crossview image pairs with accurate Geo-tags to evaluate Geo-localization performance (denoted as CVACT_test).
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types, or memory amounts) used for running its experiments. It only mentions a 'GPU gift donated by NVIDIA Corporation' in the acknowledgements, but not for the experimental setup.
Software Dependencies No The paper mentions using a 'VGG16 model' and 'Adam optimizer [7]' but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes Similar to [5, 11], we set γ to 10 for the triplet loss. Our network is trained with Adam optimizer [7], and the learning rate is set to 10 5. Exhaustive mini-batch strategy [19] is utilized to create triplet images within a batch, and the batch size Bs is set to 32.