DM-RSC

[TCSVT]: Multi-Frame Rolling Shutter Correction with Diffusion Models

1. Data Preparation

To train and test the models, please first download the datasets:

• Download Fastec and Carla datasets from the repositories:
  • DeepUnrollNet
• and download the BSRSC dataset from the same repository above.
  • BSRSC

---

3. How to Test

Download Pretrained Checkpoints

Download the pretrained model checkpoints from Google Drive.

Sampling from MDM

Run the following command to generate samples from the MDM model:

python -m script.sample_mdm --config configs/MDM/fastec_sample-MDM.yaml

Sampling from ODM

Run the following command to generate samples from the ODM model:

python -m script.sample_odm --config configs/ODM/fastec_sample_ODM.yaml

3. How to Train

Train MDM Model

Run the following command to train the MDM model:

python -m script.train_mdm --config DM-RSC/configs/MDM/fastec_train_MDM.yaml

Train ODM Model

Run the following command to train the ODM model:

python -m script.train_odm --config DM-RSC/configs/ODM/fastec_train_ODM.yaml

4. Supplementary Materials

Download supplementary materials from Google Drive.

Refer to the supplementary.zip archive for comparison GIF and additional results.