The HDR Reconstruction from a Single Raw Image track starts now! We release testing data. Check out the "Get data" page and prepare the submission!
2024.02.20 Challenge site online
2024.03.01 Validation server online
2024.04.23 Final test data release (inputs only)
2024.04.30 Test submission deadline
2024.05.05 Fact sheets and code/executable submission deadline
The dynamic range of real-world scenes frequently exceeds the capture capabilities of standard consumer camera sensors, often resulting in loss of detail in both overly bright and dark areas. To address this, the computational imaging community has extensively explored High Dynamic Range (HDR) imaging, which records a broader spectrum of intensity levels and captures more scene information. Unlike conventional Low Dynamic Range (LDR) images, HDR preserves greater detail in both over- and under-exposed areas. This enhancement not only benefits various vision tasks, such as segmentation and object detection but also produces more visually pleasing images — a goal long pursued by computer vision researchers.
To further push the research of HDR reconstruction forward, we are launching a challenge centered on reconstructing HDR images from single raw images. This approach specifically focus on single raw image HDR reconstruction, which avoids potential misalignments that can occur in multi-image fusion. We will use the Raw-to-HDR dataset called SRHDR, mainly curated by Prof. Fu’s team in [a]. This dataset contains paires of LDR and HDR images. The LDR input is captured under challenging lighting conditions, representing the over- and under-exposed regions of a high dynamic range scene. The corresponding ground truth HDR images in the dataset are produced through bracketed exposures of each scene, subsequently merged using basic HDR fusion algorithms (Debevec etal., 2008). We will host the competition using open source online platform, e.g. CodaLab. All submissions are evaluated by our script running on the server and we will double check the results of top-rank methods manually before releasing the final test-set rating.
The train and validation data is already made available to the registered participants.
Please check the terms and conditions for further rules and details.
Please test the submission process through the validation (val) phase to ensure smooth submission. If there are any bugs, please contact us promptly during the validation phase. Once the testing phase begins, to ensure fairness, we will no longer address issues related to result submissions. You can contact us by sending an email to organizers pbdl.ws@gmail.com with title 'HDR Reconstruction from a Single Raw Image Inquiry'.
We employ the standard Peak Signal To Noise Ratio (PSNR) and the Structural Similarity Index (SSIM) in grayscale, as is commonly used in the literature. The final evaluation metric can be calculated using the following formula:
$$Score=\log_k(SSIM*k^{PSNR})+\log_k(MS-SSIM*k^{PSNR-µ})=PSNR+\log_k(SSIM)+PSNR-µ+\log_k(MS-SSIM)$$
In our implementation, $k=1.2$.
For the final ranking, we will use the average Score as the primary measure. The complexity of the algorithm will only serve as a reference and will not be included in the final metric. Please refer to the evaluation function in the 'evaluate.py' of the scoring program.
The HDR Reconstruction from a Single Raw Image Challenge is one track of PBDL-Challenge, Physics Based Vision meets Deep Learning Workshop 2024, in conjunction with CVPR 2024. Participants are not restricted to train their algorithms only on the provided dataset. Other PUBLIC dataset can be used as well. Participants are expected to develop more robust and generalized methods for low light image enhancement in real-world scenarios.
When participating in the competition, please be reminded that:
Before downloading and using the dataset, please agree to the following terms of use. You, your employer and your affiliations are referred to as "User". The organizers and their affiliations, are referred to as "Producer".
@inproceedings{zou2023rawhdr,
title={Rawhdr: High dynamic range image reconstruction from a single raw image},
author={Zou, Yunhao and Yan, Chenggang and Fu, Ying},
booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
pages={12334--12344},
year={2023}
}
Industry and research labs are allowed to submit entries and to compete in both the validation phase and the final test phase. However, in order to get officially ranked on the final test leaderboard and to be eligible for awards the reproducibility of the results is a must and, therefore, the participants need to make available and submit their codes or executables. All the top entries will be checked for reproducibility and marked accordingly.
Start: Feb. 20, 2024, midnight
Description: The online evaluation results must be submitted through this CodaLab competition site of the Challenge.
Start: April 23, 2024, midnight
Description: The online evaluation results must be submitted through this CodaLab competition site of the Challenge.
April 30, 2024, 11:59 p.m.
You must be logged in to participate in competitions.
Sign In