Help TT Lab see the table
When you have a minute, place eight points on the table in a photo. One photo takes about a minute. The labels you add are used to train the table detection model of the TT Lab scoreboard.
Loading a photo…
How to label
- Look only at the one table in play closest to the middle of the frame.
- All eight points lie on the table surface. Four yellow points are the table corners. Two red points are the net ends — not the top of the net posts, but the points on the table surface where the net meets the side lines (the posts stand outside the table). Two blue points are the centre line ends — the middle of each end line, where the white centre line meets it. Order within a colour doesn't matter.
- Pick a point and tap the photo to place it; the next point is picked for you. Drag placed points to move them. A magnifier appears while you press. If the table is small, use the 1× button at the bottom right of the photo to zoom 2× or 3×, and drag an empty spot to pan.
- If a point is hidden but you can tell where it is, place it there and choose “Hidden”. If a corner is outside the picture (including black bars), don't mark it at the edge — choose “Off screen”. Don't guess positions.
- If there is no table, choose “No table”. If you can't tell, choose “Unsure” to skip.
- Keyboard: 1–8 pick a point, arrows move 1px (Shift 5px), O hidden, X off screen, Enter save and next.
You can practise without signing in. Practice points are not saved.
Good to know
- The labels you add are used to train the table detection model of the TT Lab scoreboard. We want it to find tables in amateur club footage, where it often fails today.
- Non-commercial: TT Lab is currently a non-commercial project that takes no money. Labels are used only to train the table detection model.
- Label contributions are counted anonymously. Nobody else can see who labelled what; the page shows only how many people took part and how many photos are finished. Your own count is shown only to you.
- We don't collect any new personal data. Your sign-in is used only so the same person doesn't get the same photo twice. You can delete all your labels on this page at any time.
- Each photo goes to at least three different people, and the answer is the median of each point. If people disagree a lot, the photo is left out of training and checked separately.
- Photos are frames from YouTube videos published under CC BY (Attribution). Each photo shows its channel and a link to the original. Players and channels are not affiliated with TT Lab.