Arbicon-Net: Arbitrary Continuous Geometric Transformation Networks for Image Registration

Authors: Jianchun Chen, Lingjing Wang, Xiang Li, Yi Fang

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we carried out a set of tests under different experimental settings to validate the performance of our proposed Arbicon-Net for its capability of estimating the geometric transformation for image dense correspondence in semantic alignment.
Researcher Affiliation Academia Jianchun Chen NYU Multimedia and Visual Computing Lab New York University Brooklyn, NY 11201 jc7009@nyu.edu Lingjing Wang NYU Multimedia and Visual Computing Lab New York University Brooklyn, NY 11201 lw1474@nyu.edu Xiang Li NYU Multimedia and Visual Computing Lab New York University Brooklyn, NY 11201 xl845@nyu.edu Yi Fang NYU Multimedia and Visual Computing Lab New York University Abu Dhabi Abu Dhabi, UAE yfang@nyu.edu
Pseudocode No The paper describes the architecture and components of Arbicon-Net, but it does not include any formal pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any statements about releasing open-source code or provide a link to a code repository.
Open Datasets Yes In our experiment, three image datasets, Pascal VOC dataset [35], PF-Pascal dataset [36] and Proposal Flow dataset [37] are used to prepare both synthesized and real image dataset for the various tests.
Dataset Splits Yes PF-Pascal contains 1351 semantically aligned image pairs from 20 semantic category of Pascal VOC dataset with a 7:3:3 training/validation/testing split.
Hardware Specification Yes The network is implemented by Py Torch framework and ran on an Nvidia GTX 1080Ti GPU.
Software Dependencies No The paper mentions "Py Torch framework" but does not specify a version number or list other software dependencies with their versions.
Experiment Setup Yes Arbicon-Net starts with the use of Res Net [34] (before conv4-23) with weight pre-trained on Image Net for local feature extraction, then followed by three 4-D convolution kernels in Section 3.2 which are of size (3, 3, 3) with channels (10, 10, 1) respectively, and end with four MLPs configured with the size (256, 256, 64, 2). In Arbicon-Net, we set the dimension of translation descriptor d AB at 256 as shown in Fig.4. For supervised Arbicon-Net, we use Adam optimizer for training with learning rate 0.001.