This repo contains the code for High-fidelity Event-Radiance Recovery via Transient Event Frequency (CVPR 2023 [Paper], [Video]), by Jin Han, Yuta Asano, Boxin Shi, Yinqiang Zheng, and Imari Sato.
We provide data samples in this link.
Please download the data samples and move them into the folder ./data_samples/
. The recovered radiance values will be saved in ./data_samples/xxx/ev_radiance_360x640_len4.npy
.
For hyperspectral reconstruction, since there are multiple event files captured under different light sources (with different narrow-band wavelengths), it takes longer time to get the radiance values in different wavelengths.
-
For hyperspectral reconstruction:
python TEF.py -m hyperspectral -i data_samples/painting
Then relight using different lighting files:
python relight.py -i data_samples/painting
-
For depth sensing:
python TEF.py -m depth -i data_samples/depth
-
For iso-depth contour reconstruction:
python TEF.py -m iso-contour -i data_samples/shape
If you find the paper is useful for your research, please cite our paper as follows:
@InProceedings{Han_2023_CVPR,
author = {Han, Jin and Asano, Yuta and Shi, Boxin and Zheng, Yinqiang and Sato, Imari},
title = {High-Fidelity Event-Radiance Recovery via Transient Event Frequency},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2023},
pages = {20616-20625}
}