NTIRE 2025 dataset is crafted to benchmark low-light image enhancement algorithms. It contains 219 training images and 46 validation images, each with paired low-light inputs and ground truth images. The dataset was created by author 'okhater' and last updated on April 4, -2025.
Use Cases
- Benchmarking low-light image enhancement algorithms based on the paired low-light and ground truth images.
- Training supervised learning models for image restoration based on the 219 training image pairs.
- Evaluating model generalization on unseen low-light conditions based on the separate 46-image validation set.
Strengths
- Dataset is explicitly structured for both development and evaluation phases of a challenge.
- Contains 219 paired low-light and ground truth images for training.
- Includes a separate validation set of 46 low-light images for evaluation.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count for the full dataset is unknown, which may limit suitability assessment.
Provenance
- Source
- huggingface
- Collection Method
- Crafted for the NTIRE 2025 Low Light Image Enhancement Challenge.
- Time Range
- 2025
- Freshness
- Last updated 2025-04-04 00:15:11; freshness should be verified.
- Geography
- null