Evaluation Metric
In the challenge, we utilize three evaluation metrics for assessment: Root Mean Square Error (RMSE), Structural Similarity (SSIM), and Peak signal-to-noise ratio (PSNR).
\(Score=\log_k(SSIM*k^{PSNR})-RMSE=PSNR+\log_k(SSIM)-RMSE\)
In our implementation, k=1.2.
Format of Submission
To submit your results, you must provide a .zip file. The name of the zip file can be chosen at your discretion.
The training data are structured in the following hdf5 format:
And the val/test data are structured as:
And as your submission, you should upload a zip file which are as the following format:
Format of Filenames
Within the zip file, each video sequence's reconstruction folder should be named identically to its input name. For instance, the HDR reconstruction result for the input named：
1.h5
should be placed in a folder named
< 1 >
The reconstructed video frames for "1.h5" must be located inside the "1" folder, with names following a five-digit zero-padded format, such as
1/00000.png, 1/00001.png, ……, 1/00399.png
All the png files should be under the same folder, there should be no subfolders in your submission.
Format of Images
Your reconstruction frames should be saved as PNG images, using the uint16, single-channel, PNG format.
Submission