quintonqu / esim-ad

Event-based Camera Simulation using Monte Carlo Path Tracing with Adaptive Denoising

Event-based Camera Simulation using Monte Carlo Path Tracing with Adaptive Denoising

Y. Tsuji, Y. Yatagawa, H. Kubo, and S. Morishima, "Event-based Camera Simulation using Monte Carlo Path Tracing with Adaptive Denoising," In Proceedings of the IEEE International Conference on Image Processing, 2023. (to appear) [Preprint]

Install

Prepare virtualenv using Poetry.

# Install required modules
poetry install --no-dev

# Enable Poetry's virtualenv
poetry shell

Dataset

• SanMiguel: https://drive.google.com/file/d/1B8Juxd1llp__96dWlXDegZ-YuBRj27rc/view?usp=share_link
• TwoBoxes: https://drive.google.com/file/d/1XWpIgi9oR5SpYOAm_1CbnvixrfL42HdK/view?usp=drive_link
• LivingRoom: https://drive.google.com/file/d/1prC6BAMmmqR0TJklhVbuiB3BnrlXJS4y/view?usp=drive_link

Then, unzip the archive and store in the data folder. If the data is sanmiguel, four subdirectories (i.e., 32spp, 64spp, 128spp, and 4096spp) will be stored in data/sanmiguel.

Run

# To reproduce our results in the paper, run the following.
python main.py --data_root /path/to/data --spp 32 --ksize 13 --wlr simple --thres 1.00

The default parameters used to generate the following results are shown in run.sh. After running the above code, you can visually compare our method with other baselines with analysis.ipynb in the notebooks folder.

Results

Living Room (32spp)

LivingRoom-comp.mp4

San Miguel (32spp)

SanMiguel-comp.mp4

Two Boxes (32spp)

TwoBoxes-comp.mp4

Note that the above videos are compressed to fit the 10MB file size limit of GitHub. You can find uncompressed ones in the following URL:
https://drive.google.com/file/d/1hPtCmBYmo7h-aczLSPVK8ZIK8_sUJ4nW/view?usp=share_link

Also, you can find the quantitative comparisons in the supplementary document in the following URL:
https://drive.google.com/file/d/1gJsuApMHOO-PPWweZ849QaGdr0OQsjyl/view?usp=share_link

License

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

(c) Yuta Tsuji, Tatsuya Yatagawa, Hiroyuki Kubo, and Shigeo Morishima.