Shay85Gil/somethings_missing_here

Something's Missing Here

A dataset of viewpoint-aligned before/after image pairs for training and evaluating missing-object detection: each pair shows the same scene from (nearly) the same viewpoint, once with one or more objects present and once after they were removed, with a label naming what went missing.

The dataset originated from pairs derived from the simkoc/Remove360 dataset (training/Remove360_based/), extended with original photographs collected by the dataset author (training/DeTaken/), plus a held-out test/ split of scenes that appear nowhere in training.

Structure

training/
├── DeTaken/                 original photographs by the dataset author
│   └── <scene>/             boxes, cabinet_window, magnets, red_stool, ...
│       └── pair_<nn>/
│           ├── before.jpg   earlier image
│           ├── after.jpg    later image
│           └── label.json   {"missing": true, "items": ["<removed object>"]}
│                            or {"missing": false, "items": []} for no-change pairs
└── Remove360_based/         pairs derived from Remove360 (see below)
    └── <scene>/             backyard, bedroom, living-room, office, park, stairwell
        └── pair_<nn>/       same pair layout
test/
└── test_<scene>/            held-out scenes, one pair each, never in training
    └── pair_01/             same pair layout

Overview:

split	collection	scenes	pairs	positive	negative
training	DeTaken	13	87	50	37
training	Remove360_based	6	146	124	22
test	—	7	7	7	0

Remove360-derived negative (no-change) pairs additionally carry a source field in label.json documenting which images they were built from.

In Remove360_based/ pairs, before.jpg is warped into after.jpg's camera frame and both images are cropped to their shared valid region, so the two images are pixel-aligned with identical dimensions. DeTaken/ and test/ pairs are handheld re-shots from approximately the same viewpoint and are not pixel-aligned.

The Remove360_based subset

Remove360 provides separate pre-removal and post-removal camera walks of real indoor and outdoor scenes. Its before and after images are independent captures — they are not pixel-aligned pairs — so this subset was built by finding and aligning the closest matching viewpoints between the two walks.

146 pairs — 124 positives (something was removed) and 22 negatives (nothing changed) — across 6 scenes and 9 removed objects:

scene	positive	negative		removed object	pairs
backyard	53	7		chairs	36
stairwell	24	7		backpack	24
living-room	21	5		stroller	16
office	13	1		sofa	15
park	10	2		deckchair	11
bedroom	3	0		bicycle	10
				pillows	6
				table	3
				toy-truck	3

How it was generated

1. Download — the full simkoc/Remove360 repository (file tree of <scene>/<object>/train|test|masks), where train/ holds pre-removal ("before") images and test/ holds post-removal ("after") images.
2. Valid-region cropping — a subset of Remove360's images is truncated at fixed byte boundaries on the Hub itself (all of backyard/stroller at 2.75 MiB, all of backyard/playhouse at 256 KiB); truncated JPEGs decode with a uniform gray tail. Each image was cropped to its real content before matching, and images with less than 15% real content were discarded.
3. Viewpoint matching — every after image was ranked against all before images of the same object by SIFT feature matches (Lowe ratio 0.75); the top 3 candidates were verified with a RANSAC homography (reprojection threshold 4 px, minimum 40 inliers).
4. Acceptance criteria — a pair was kept only if each frame covers at least 85% of the other under the homography (mutual frame coverage) and the warped before image correlates with the after image at ≥ 0.475 zero-mean normalized correlation. Each before image was used in at most one pair.
5. Alignment and cropping — the accepted before image was warped into the after frame at full resolution and both images were cropped to the largest rectangle of shared valid pixels.
6. Labeling — each pair's label.json records the removed object (the Remove360 object folder the pair came from) as {"missing": true, "items": ["<object>"]}.
7. Manual curation — the automatically accepted pairs were reviewed and some were deleted by hand; the remaining pairs were renumbered contiguously.
8. No-change negatives — within one scene, the before images of two different pairs that share the same removed-item label come from the same pre-removal camera walk (the item is present in both), and likewise the after images from the same post-removal walk (absent in both). Such same-side image pairs were matched and aligned with the same gates as the positives — with mutual frame coverage additionally capped at 96% so overlap statistics cannot separate the classes — and saved as {"missing": false} pairs.

Known limitations

• No playhouse pairs — all of Remove360's backyard/playhouse images are truncated to ~6% of their content on the Hub, which is below the usability floor.
• Stroller pairs are half-height — backyard/stroller images are truncated to roughly their top half, so its pairs are wide bands (3900×1000) rather than full frames (4000×2200).
• Residual parallax — alignment uses a single homography per pair; small parallax between the two camera positions can remain, especially on close foreground geometry.
• Some objects other than the labeled one may have shifted slightly between Remove360's two capture sessions.

License and attribution

The Remove360_based/ subset is a derivative of simkoc/Remove360 and is distributed under the same CC-BY-NC-4.0 license, which this dataset adopts as a whole. If you use it, please also cite the original Remove360 paper (arXiv:2508.11431).