In this competition, your task is to develop models that allow recreation of reflectance measurements. For this task, partial reflectance measurements are provided as a dataset of images, along with calibration data that is necessary to evaluate analytical reflectance models. The goal is, given the known reflectance and calibration data, to recreate unseen reflectance given only its corresponding calibration data. The model outputs are submitted as data files and are automatically rated and ranked according to several metrics.
Please see the provided starting kit for an example evaluation script under "Participate" -> "Files".
The model outputs need to be uploaded in the same format as the reflectance images in the dataset download, i.e. in the OpenEXR file format with the same naming scheme. They are then compared to the hold-out set of reflectance images unknown to challenge participants. The evaluation makes use of multiple common image metrics like RMSE, MAE, SSIM, as well as metrics like HDR-VDP that are tailored to provide difference ratings on high dynamic range images that resemble judgements by a human observer. Metric scores are averaged over all pixels and all images.
Start: May 15, 2020, midnight
You must be logged in to participate in competitions.Sign In