Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
An Authentic Dataset for Visual-Inertial Odometry
| Date | Stars |
|---|---|
| 2026-07-24 | 269 |
| 2026-07-25 | 269 |
| 2026-07-28 | 269 |
| 2026-07-30 | 269 |
| 2026-08-06 | 269 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
 # ADVIO: An Authentic Dataset for Visual-Inertial Odometry [Santiago Cortés](https://research.aalto.fi/portal/santiago.cortesreina.html) · [Arno Solin](http://arno.solin.fi) · [Esa Rahtu](http://esa.rahtu.fi) · [Juho Kannala](https://users.aalto.fi/~kannalj1/) The lack of realistic and open benchmarking datasets for pedestrian visual-inertial odometry has made it hard to pinpoint differences in published methods. Existing datasets either lack a full six degree-of-freedom ground-truth or are limited to small spaces with optical tracking systems. We take advantage of advances in pure inertial navigation, and develop a set of versatile and challenging real-world computer vision benchmark sets for visual-inertial odometry. For this purpose, we have built a test rig equipped with an iPhone, a Google Pixel Android phone, and a Google Tango device. We provide a wide range of raw sensor data that is accessible on almost any modern-day smartphone together with a high-quality ground-truth track. We also compare resulting visual-inertial tracks from Google Tango, ARCore, and Apple ARKit with two recent methods published in academic forums. The data sets cover both indoor and outdoor cases, with stairs, escalators, elevators, office environments, a shopping mall, and metro station. ## Example video This video shows the ground-truth track for data set \#16 (captured in one of the two office buildings). The visualized track is the ground-truth track calculated from the entire IMU data sequence. The fix points used for track calculation are visualized by dots. The track on the current floor shows in red. The video has been sped-up. [](http://www.youtube.com/watch?v=AU_PXxvxBHM) ## Attribution If you use this data, please cite the original paper presenting it: * Santiago Cortés, Arno Solin, Esa Rahtu, and Juho Kannala (2018). *ADVIO: An authentic dataset for visual-inertial odometry.* Accepted for publication in European Conference on Computer Vision (ECCV). Munich, Germany. [[arXiv]](https://arxiv.org/abs/1807.09828) # Downloading the data The data files are available for download on Zenodo: [https://zenodo.org/record/1320824](https://zenodo.org/record/1320824) and can be downloaded on a per dataset basis from there. You can also use `wget` with the following bash snippet to fetch all the data: ```bash # Download all 23 data ZIPs from Zenodo for i in $(seq -f "%02g" 1 23); do wget -O advio-$i.zip https://zenodo.org/record/1476931/files/advio-$i.zip done ``` The size of one set ranges from 71 Mb to 255 Mb (packed). Uncompressed total size around 5.1 Gb. # Details on collected data  ## Ground-truth * __Ground-truth poses:__ Camera pose (translation and orientation) calculated based on the raw IMU data and a set of known fixation points. The ground-truth track is sampled at 100 Hz. * __Fix points:__ A set of ground-truth points marked with a visual editor. The points are based on the three videos stored by the system (primarily the iPhone and the second iPhone that filmed a reference track showing the capturer) and floor plan layouts. ## iPhone * __Camera frames:__ Camera frames are captured at 60 fps (resolution of 1280 by 720, portrait). The exact frame acquisition times reported by the platform are stored. The frames are packed into an H.264/MPEG-4 video file. * __Platform location:__ Data collected through CoreLocation. The update rate depends on the device and its capabilities. Locations are requested with the desired accuracy of kCLLocationAccuracyBest. The timestamps are converted to follow the same clock as the other sensors (time interval since device boot). The stored values are * Coordinate.latitude * Coordinate.longitude * HorizontalAccuracy * Altitude * VerticalAccuracy * Speed
Excerpt of 8,981 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:128763e62b9ae338, topic:computer-vision, readme:computer vision