CoMIR: Contrastive Multimodal Image Representation for Registration

Authors: Nicolas Pielawski, Elisabeth Wetzer, Johan Öfverstedt, Jiahao Lu, Carolina Wählby, Joakim Lindblad, Natasa Sladoje

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We assess the extent of achieved rotational equivariance and the stability of the representations with respect to weight initialization, training set, and hyperparameter settings, on a remote sensing dataset of RGB and near-infrared images. We evaluate the learnt representations through registration of a biomedical dataset of bright-field and second-harmonic generation microscopy images; two modalities with very little apparent correlation.
Researcher Affiliation Academia Nicolas Pielawski , Elisabeth Wetzer*, Johan Öfverstedt, Jiahao Lu, Carolina Wählby, Joakim Lindblad, Nataša Sladoje Dept. of Information Technology, Uppsala University, Sweden
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code is available at: https://github.com/MIDA-group/Co MIR.
Open Datasets Yes Zurich Dataset. The open Zurich dataset [57] consists of 20 aerial images of the city of Zurich of about 930 940px.
Dataset Splits Yes The training set consists of 40 image pairs of size 834 834px which are center-cropped from the corresponding original images. The validation set for the CNN training consists of another 25 such pairs, a tuning set for registration parameters of additional 7 pairs, and the test set for evaluation of another 134 pairs.
Hardware Specification No The paper mentions 'training (GPU, 1345s)' and 'inference (CPU, 5s/image)' but does not provide specific hardware details such as GPU or CPU models.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers, such as programming language versions or library versions (e.g., Python 3.8, PyTorch 1.9).
Experiment Setup Yes We set the temperature to τ = 0.5. We use 46 negative samples. ... For both datasets patches of the size 128 128px were chosen. ... The SHG images were preprocessed by applying a log-transform log(1 + x) with x [0, 1].