Reflection Separation using a Pair of Unpolarized and Polarized Images

Authors: Youwei Lyu, Zhaopeng Cui, Si Li, Marc Pollefeys, Boxin Shi

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Experimental ResultsWe evaluate our method on both synthetic and real data with extensive experiments including the comparison with related work and ablation study. For all quantitative evaluations, both the peaksignal-to-noise ratio (PSNR) and the structural similarity index measure (SSIM) are used to evaluate the quality of separated images.
Researcher Affiliation Academia 1Beijing University of Posts and Telecommunications 2Department of Computer Science, ETH Zürich 3National Engineering Laboratory for Video Technology, Peking University 4Peng Cheng Laboratory
Pseudocode No The paper describes the network architecture and physical model using text and equations but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes The code and test data are available at https://github.com/Youwei Lyu/reflection_separation_with_un-polarized_images.
Open Datasets Yes At the first step, we randomly pick two images from PLACE2 dataset [26] as original reflection and transmission layers.
Dataset Splits No We use 5000 pairs of images from our synthetic validation dataset with ground truth reflection and transmission layers to quantitatively compare our method with state-of-the-art approaches.
Hardware Specification No The paper mentions using a “Lucid Vision Phoenix polarization camera” for data capture but does not provide specific hardware details (GPU/CPU models, memory) used for running the experiments or training the models.
Software Dependencies No We implement our model using Py Torch deep learning framework [17]. Adam [11] is used as the optimizer with a starting learning rate of 0.0004, β1 = 0.9 and β2 = 0.999.
Experiment Setup Yes Adam [11] is used as the optimizer with a starting learning rate of 0.0004, β1 = 0.9 and β2 = 0.999. The learning rate is descended to 0.0002 and 0.00008 after 12th and 18th epochs respectively. λ1,2,3,4 are set to be 1.2, 1.5, 1.0, and 1.5 respectively for our training.