Cross-View Geo-Localization via Learning Disentangled Geometric Layout Correspondence

Authors: Xiaohan Zhang, Xingyu Li, Waqas Sultani, Yi Zhou, Safwan Wshah

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments show that Geo DTR not only achieves state-of-the-art results but also significantly boosts the performance on same-area and cross-area benchmarks. To evaluate the effectiveness of Geo DTR, we conduct extensive experiments on two datasets, CVUSA (Workman, Souvenir, and Jacobs 2015), and CVACT (Liu and Li 2019).
Researcher Affiliation Academia 1Department of Computer Science, University of Vermont, Burlington, USA 2Vermont Complex Systems Center, University of Vermont, Burlington, USA 3 Shanghai Center for Brain Science and Brain-Inspired Technology, China 4 Intelligent Machine Lab, Information Technology University, Pakistan 5 NEL-BITA, School of Information Science and Technology, University of Science and Technology of China, China
Pseudocode No The paper provides architectural diagrams and mathematical equations but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes Our code can be found at https://gitlab.com/vail-uvm/geodtr.
Open Datasets Yes To evaluate the effectiveness of Geo DTR, we conduct extensive experiments on two datasets, CVUSA (Workman, Souvenir, and Jacobs 2015), and CVACT (Liu and Li 2019).
Dataset Splits Yes Both CVUSA and CVACT contain 35, 532 training pairs. CVUSA provides 8, 884 pairs for testing and CVACT has the same number of pairs in its validation set (CVACT val). Besides, CVACT provides a challenging and large-scale testing set (CVUSA test) which contains 92, 802 pairs.
Hardware Specification Yes We train the model on a single Nvidia V100 GPU for 200 epochs with Adam W (Loshchilov and Hutter 2017) optimizer.
Software Dependencies No The paper mentions "Adam W (Loshchilov and Hutter 2017) optimizer" but does not specify version numbers for any key software components or libraries (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes α and β are set to 10 and 5 respectively. We train the model on a single Nvidia V100 GPU for 200 epochs with Adam W (Loshchilov and Hutter 2017) optimizer.