Featuring video sequences and depth maps extracted from high-definition SlowTV broadcasts, such as long-form train journeys and walking tours. It serves as the training and evaluation data for the ICCV2023 paper 'Kick Back & Relax', focusing on monocular depth reconstruction from unconstrained video streams.
Use Cases
- Train self-supervised monocular depth models using the sequential video frames to exploit temporal consistency.
- Evaluate the performance of depth estimation algorithms in diverse, real-world outdoor settings.
- Benchmark structure-from-motion (SfM) pipelines on long-duration, high-resolution scenic video sequences.
Strengths
- Derived from high-definition SlowTV video sources featuring continuous, stable camera motion across diverse landscapes.
- Includes data used to benchmark the ICCV2023 monocular depth reconstruction framework.
- Covers a wide variety of outdoor environments including rail journeys, urban walking tours, and boat trips.