Dynamic Anchor Learning for Arbitrary-Oriented Object Detection

Authors: Qi Ming, Zhiqiang Zhou, Lingjuan Miao, Hongwei Zhang, Linhao Li2355-2363

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on three remote sensing datasets HRSC2016, DOTA, UCAS-AOD as well as a scene text dataset ICDAR 2015 show that our method achieves substantial improvement compared with the baseline model.
Researcher Affiliation Academia Qi Ming, Zhiqiang Zhou , Lingjuan Miao, Hongwei Zhang, Linhao Li School of Automation, Beijing Institute of Technology, China chaser.ming@gmail.com, {zhzhzhou, miaolingjuan}@bit.edu.cn, zhanghw.hongwei@gmail.com, lilinhao@bit.edu.cn
Pseudocode No The paper describes its method in prose and mathematical equations but does not provide a formal pseudocode block or algorithm.
Open Source Code Yes The code and models are available at https://github.com/ming71/DAL.
Open Datasets Yes We conduct experiments on the remote sensing datasets HRSC2016, DOTA, UCAS-AOD and a scene text dataset ICDAR 2015. The ground-truth boxes in the images are all oriented bounding boxes. The HRSC2016 (Liu et al. 2017b) is a challenging remote sensing ship detection dataset... DOTA (Xia et al. 2018) is the largest public dataset... UCAS-AOD (Zhu et al. 2015) is an aerial aircraft and car detection dataset... The ICDAR 2015 dataset... (Karatzas et al. 2015).
Dataset Splits Yes The HRSC2016 (Liu et al. 2017b) is a challenging remote sensing ship detection dataset, which contains 1061 pictures. The entire dataset is divided into training set, validation set and test set, including 436, 181 and 444 images, respectively. ... UCAS-AOD (Zhu et al. 2015) is an aerial aircraft and car detection dataset, which contains 1510 images. We randomly divide it into training set, validation set and test set as 5:2:3.
Hardware Specification Yes We train the models on RTX 2080Ti with batch size set to 8.
Software Dependencies No The paper mentions software components like "Retina Net", "Res Net-50", "FPN", and "Adam" but does not provide specific version numbers for any of these or other software libraries/frameworks.
Experiment Setup Yes The initial learning rate is set to 1e-4 and is divided by 10 at each decay step. The total iterations of HRSC2016, DOTA, UCAS-AOD, and ICDAR 2015 are 20k, 30k, 15k, and 40k, respectively. We train the models on RTX 2080Ti with batch size set to 8. All images are resized to 800 800. ... anchors are set with aspect ratios of {1/2, 1, 2}.