Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
image
imagewidth (px)
1.02k
1.02k
label
class label
2 classes
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
1prompt_2
1prompt_2
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
0prompt_1
1prompt_2
1prompt_2
0prompt_1
0prompt_1

SAPset global edit cases

The 12 whole-object edits on which SAPset's global editing was evaluated: 10 source objects of Edit3D-Bench, each with one or two edited views that change the whole object (made of doughnuts, rock, glass, fur, spikes, a hexagonal pattern, sausages, an orc, yarn, dense and sparse wire mesh).

Layout

The folders follow Edit3D-Bench's layout, without the source models:

metadata.json                               objects and their edit descriptions (prompt_1, prompt_2)
<dataset>/<object>/prompt_<k>/2d_edit.png       the edited view (1024 x 1024)
<dataset>/<object>/prompt_<k>/2d_mask.png       the 2D edit mask: a box around the whole object
<dataset>/<object>/prompt_<k>/3d_edit_region.glb  the 3D edit region: a box around the whole object
<dataset>/<object>/prompt_<k>/annotation_view.json  which of the benchmark's 16 cameras the edited view uses

Use

With the SAPset code repository and a local copy of Edit3D-Bench:

python scripts/download_global_cases.py --edit3d-bench Edit3D-Bench/data --out global_edits

downloads these files and links each object's source_model/ from Edit3D-Bench, so that --bench-root global_edits works with every SAPset stage. The repository's README lists the commands that reproduce the published global results: 10 cases with the local configuration and this whole-object region, and the two wire-mesh pumpkin cases with one uniform adapter strength of 0.5.

The per-object adapters those commands use are in Itan-levin/sapset-adapters.

License

To be added.

Citation

To be added.

Downloads last month
84