Real-time depth estimation is a subject of significant importance with far-reaching implications across various fields such as Augmented/Virtual Reality, Autonomous Navigation, Robotics and more. This technology holds the potential to revolutionize several domains, making it a vital area of research and application.
On this repository, we'll learn how to use MiDaS to estimate the depth of a room in "real-time" (we'll use a video to emulate a real-time performance and we'll check that the number of fps is adequate) with a monocular camera (a camera with just one objective).
- A jupyter notebook, containing all the necessary code to load and execute MiDaS on our example (be careful, the pretrained_weight of the medium version of MiDaS have a size of 500MB).
- A data folder, containing the video whoose depth we're going to estimate.
- The output video in .mp4 format.
Victor Hoffmann