MODEST: Ultra-High-Resolution Multi-Optics Depth-of-Field Stereo Dataset
A team of researchers has unveiled MODEST, the inaugural ultra-high-resolution (20MP, 5472x3648 pixels) multi-optics depth of field stereo DSLR dataset. This dataset aims to fill the gap of large-scale, full-frame, high-fidelity real-image datasets essential for training and assessing computer vision algorithms in shallow depth of field (DoF) rendering and defocus deblurring. It features 20,000 images captured across 50 unique optical configurations, varying focal length and aperture. Such systematic variation is vital, as the optical effects of shallow DoF and defocus blur are closely linked to camera settings, necessitating thorough model evaluations. MODEST supports contemporary applications like AR, VR, smartphones, and industrial robots. Detailed information can be found in a paper on arXiv (arXiv:2511.20853).
Key facts
- MODEST is the first ultra-high-resolution (5472x3648px, 20MP) multi-optics depth of field stereo DSLR dataset.
- The dataset includes 20,000 images across 50 distinct optical configurations.
- It systematically varies focal length and aperture for complex real-world scenes.
- The dataset addresses the lack of large-scale, full-frame, high-fidelity real-image datasets for shallow DoF rendering and defocus deblurring.
- Optical effects of shallow DoF and defocus blur depend on camera optical configuration (focal length and aperture).
- The dataset supports applications like AR, VR, smartphones, and industrial robots that use stereo or multi-camera systems.
- MODEST captures the optical realism and complexity of professional camera systems.
- The paper is available on arXiv with identifier arXiv:2511.20853.
Entities
Institutions
- arXiv