A Plug-and-Play Image Registration Network

Authors: Junhao Hu, Weijie Gan, Zhixin Sun, Hongyu An, Ulugbek Kamilov

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our numerical results on OASIS and CANDI datasets show that our methods achieve state-of-the-art performance on DIR.
Researcher Affiliation Academia Junhao Hu , Weijie Gan , Zhixin Sun, Hongyu An, Ulugbek S. Kamilov Washington University in St. Louis, St. Louis, MO, USA {hjunhao,weijie.gan,zhixin.sun,hongyuan,kamilov}@wustl.edu
Pseudocode No No explicit pseudocode or algorithm block was found.
Open Source Code No The paper does not provide an explicit statement or link to open-source code for the described methodology.
Open Datasets Yes We validated PIRATE and PIRATE+ on two widely used datasets: OASIS-1 (Marcus et al., 2007) and CANDI (Kennedy et al., 2012).
Dataset Splits No For both datasets, we randomly shuffled the images and allocated 100 unique image pairs for training and another 100 unique image pairs for evaluation. No explicit mention of a separate validation split or its size was found.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types) used for running its experiments.
Software Dependencies No The paper mentions software components like 'Adam optimizer' and 'Dn CNN' but does not provide specific version numbers for any software, libraries, or programming languages used.
Experiment Setup Yes In training phase of the denoiser, we used Adam (Kingma & Ba, 2014) optimizer with learning rate 1e 4 for 400 epochs... For DEQ in PIRATE+, we used Adam optimizer with learning rate 1 10 5 for 50 epochs. We assigned w0 to 1, w1 to 5, w2 to 1 for both datasets... PIRATE achieved the best performance by assigning γ0 to 5 105, α to 5 10 1, and τ to 1 10 7 for OASIS-1. α was assigned to 5 10 1 for CANDI while γ0 and α stayed the same.