Damian P
Member
Hi all,
This is something that I've been working on for years, and I honestly didn't think I'd get to this point, but here we are.
Cube-Vision is an app that records your solves, runs a vision algorithm to reconstruct the moves, and then compiles your stats.
You can use any cube with a normal color scheme, no bluetooth connection required.
A short demo
iOS: https://apps.apple.com/us/app/cube-vision/id6794732933
Android: Currently in closed testing. Email me at [email protected] asking for access and I can add you to the closed test.
(Android version does not have the algorithm trainer yet)
Current Features:
*The hand tracking for the timer start/stop is heavily inspired by swiftcuber.com. If you're not interested in solve reconstruction and just want to start / stop the timer by putting your hands down you should use swiftcuber instead.
Future Features (I'd love to hear other recommendations as well):
If you could share with me your feedback or any issues that you run into I'd really appreciate it.
This is something that I've been working on for years, and I honestly didn't think I'd get to this point, but here we are.
Cube-Vision is an app that records your solves, runs a vision algorithm to reconstruct the moves, and then compiles your stats.
You can use any cube with a normal color scheme, no bluetooth connection required.
A short demo
iOS: https://apps.apple.com/us/app/cube-vision/id6794732933
Android: Currently in closed testing. Email me at [email protected] asking for access and I can add you to the closed test.
(Android version does not have the algorithm trainer yet)
Current Features:
- Move by move reconstruction of 3x3 solves
- CFOP stage segmentation with per stage times, move counts, and TPS, both average and for each solve
- Recognition time for each stage
- Solve replay with the recorded timings of each move
- Average times for each OLL and PLL case
- Hand tracking for automatic starting and stopping of the timer*
- Solve export to share reconstruction + video with friends
*The hand tracking for the timer start/stop is heavily inspired by swiftcuber.com. If you're not interested in solve reconstruction and just want to start / stop the timer by putting your hands down you should use swiftcuber instead.
Future Features (I'd love to hear other recommendations as well):
- Lower extraction times (app update should be out within a week, already works in testing, should be easily faster than real time so solves won't get backed up)
- This should allow for an algorithm trainer in the near future as well
- Stage breakdowns for Roux (I might have to actually learn Roux for this one)
- Deeper analysis for Cross + F2L
- QoL changes: streamline UI, add light mode, etc.
If you could share with me your feedback or any issues that you run into I'd really appreciate it.
Last edited: