A synthetic dataset derived from MPI Sintel frames using the ELD low-light noise preset. It contains corrupted RGB frames for training and test splits, created by LForster and updated in April 2026.
Use Cases
- Evaluate optical flow model robustness using noisy RGB frames in low-light conditions.
- Fine-tune pretrained optical flow models on ELD-corrupted Sintel training frames.
- Perform controlled experiments analyzing performance degradation on synthetic low-light test frames.
Strengths
- Provides a controlled synthetic environment for low-light experiments.
- Includes separate training and test splits for model development and evaluation.
Limitations
- All data is synthetic, lacking real-world sensor noise and lighting variations.
- Specific sample size, resolution, and noise parameters are unknown.
Provenance
- Source
- Derived from the MPI Sintel optical flow dataset.
- Collection Method
- Synthetic corruption using the ELD low-light noise preset.
- Time Range
- null
- Freshness
- null
- Geography
- null