Sample project
Pixel-accurate segmentation: 18 traced masks on a single frame


- Data
- Video · 1920×1080
- Volume
- 18 masks on frame 54 · 150 frames tracked
- Status
- Sample project
The problem
Boxes tell a model roughly where an object is. Manipulation needs more: the exact edge of a knife blade, the rim of a bowl, where a hand ends and a tomato begins.
What we did
- Objects outlined with SAM 2.1 video segmentation from reviewer-placed anchors
- Every outline reviewed; drifting or leaking masks corrected
- Outlines kept in COCO segmentation format and rasterised to class masks
- Small objects (garlic cloves, herbs) drawn on top so they are never hidden by larger ones
Result
18 traced masks on this frame, 18 and 24 on the other two samples, delivered as COCO polygons and PNG class masks.
Labeled video · 30 seconds
- Caption track
- 00:00The last tomato half is sliced and the slices are added to the glass bowl.
- 00:01The board is cleared and a garlic bulb is brought over.
- 00:04Garlic cloves are broken off the bulb and peeled by hand.
- 00:12The peeled garlic is minced with the knife.
- 00:18The minced garlic is gathered by hand and sprinkled over the tomatoes.
- 00:22Salt is pinched from the salt bowl and sprinkled over the tomatoes.
- 00:26Pepper is pinched from the pepper bowl and sprinkled over the tomatoes.
- 00:28Serving tongs are brought in and the bowl is drawn to the centre of the board.
- 00:29The salad is tossed with the tongs.
Frames, depth and review sheet








Sample output
An excerpt from the delivered files. "source": "ai" marks boxes propagated by the segmentation model; every one was reviewed by a person before export.
{
"image_id": 55,
"category": "bowl stack",
"bbox": [13, 85, 217, 299],
"segmentation": [[102, 86, 83, 92, 60, 107, 35, 137, "…"]],
"area": 52605.16
}Source footage: “Farmers Market Food Demo - Spanish Tomato Salad”, US Department of Agriculture (Agricultural Marketing Service), via Wikimedia Commons (Public domain). Labels, captions, tracks and depth by Annoroid.