ApertureLab Synthetic SAS Dataset, version 1.1

Version 1.1 of the ApertureLab Synthetic SAS Dataset contains 1000 swaths from a simulated HISAS 1030-class synthetic aperture sonar, with the same 200 m by 80 m footprint and 2.5 cm resolution as version 1.0. Each swath is supplied as a georeferenced palette GeoTIFF with COCO labels in 27 categories (four target shapes and 23 kinds of clutter). It also includes a terrain height map, a table of the simulated boulders and the scene file. The dataset is published on Hugging Face under the ALSD-1.0 license.

Version 1.1 addresses four shortcomings of version 1.0. Its label polygons did not account for burial or terrain, its seafloor relief was at most 8 cm, it used a single acquisition geometry, and an information box was rendered into every image. The changes are listed below. Version 1.0 remains available and is described on its own page; the two releases are compared on the Datasets page.

Summary

Size, access and license

Size: 1000 swaths, split 800 for training, 100 for validation and 100 for testing, covering 4.66 nmi² (16 km²) of seafloor.
Hugging Face: huggingface.co/datasets/idg101/Aperture_Lab_Synthetic_Aperture_Sonar_v1_1
License: ALSD-1.0, the ApertureLab Synthetic SAS Dataset License, unchanged from v1.0.

Changes

Changes from version 1.0

The swath footprint and the sonar are unchanged: 200 m of ground range by 80 m along track at 2.5 cm, from the HISAS 1030 preset. Every scene was drawn from a new master seed. Each change is defined below.

Label geometry: Each footprint polygon is reduced according to the burial of its object, and every target and clutter object also carries a shadow polygon. Every vertex is projected through the beamformer's slant-to-ground mapping using the terrain height beneath it, so the polygon registers with the return of the object on slopes several meters high.
Label categories: Version 1.0 exported rock fields, trawl scars and pockmarks as whole-image boxes (1024 of its 5503 annotations). Version 1.1 labels objects only, in 27 categories under two supercategories. The supercategory target holds four classes: a wedge in two resting attitudes, a truncated cone and a cylinder. The supercategory clutter holds each boulder individually as rock and each of the 22 clutter shapes and meshes as its own category. Seafloor features are recorded in a per-image record instead.
Target balance: The scene plan places 2000 targets, 500 of each class.
Terrain relief: Rolling bathymetry, a regional tilt, sand waves and rock outcrops are drawn per scene and clipped to ±5 m about the scene datum; about one fifth of the scenes remain flat. Version 1.0 had at most 8 cm of small-scale relief.
Ripples: Ripple-family scenes contain one to three ripple zones with wavelengths from 0.25 m to 5 m, each with its own direction and height, and some decayed as relict ripples. Version 1.0 used a single wavelength band from 0.3 m to 1.5 m.
Acquisition: Altitude, sound speed, receiver noise and a small residual motion are drawn per scene and recorded in its label file.
Image format: The image is an 8-bit palette GeoTIFF whose pixel values are the gray display image, with nothing rendered over it.
Additional products: Each swath includes its terrain as a 16-bit height map and every simulated boulder as a row of a CSV file.
Negatives and held-out family: 300 of the 1000 scenes contain no target, compared with 150 in version 1.0, and the pockmarked seafloor family appears in every split rather than only in the test split.

Labels

Label categories

A detector trained on this set must distinguish four target shapes from one another and from every other object that returns an echo from the seafloor. The figure shows one labeled example of each target class and of two clutter categories, a boulder and a placed object, each cropped from a swath of the dataset around its COCO box.

Six sonar chips in a row, each 6 m square with its label polygon drawn and its category name below: wedge_stl, wedge_belly, truncated_cone, cylinder, a placed clutter object and a boulder (rock) — **Each target class and two clutter categories**6 m chips at the native 2.5 cm, cropped around the COCO box, with the footprint polygon solid and the shadow polygon dashed; targets in green, a placed clutter object in magenta and a boulder in orange; a 1 m bar in each corner

1 wedge_stl: This target class is a wedge resting level with its base sunk into the bed, so only its upper part stands proud.
2 wedge_belly: This target class is the same wedge lying on its flat underside; it stands higher off the bed and casts a longer shadow.
3 truncated_cone: This target class is a truncated cone standing on its base.
4 cylinder: This target class is a cylinder lying on its side.
5 rock: This clutter category labels individually every simulated boulder of 0.75 m diameter or more from the rocks CSV; smaller boulders appear in the CSV only.
6 to 17: These clutter categories are the parametric shapes pyramid, cone, frustum, dome, drum, slab, hexprism, capsule, cross, polyhedron, mound, torus.
18 to 27: These clutter categories are the meshes tire, barrel_oil, keg, sunken_anchor, fish_trap_square, ladder, debris1, ship_propeller, europalette, rock_pile.
28 clutter_other: This category is reserved for clutter kinds outside this dataset, such as a user's own mesh in an ApertureLab scene; it never occurs in v1.1.

Every target carries its footprint polygon shrunk by burial, a shadow polygon, and its bed height, grazing angle, substrate and range band as attributes. Categories 1 to 4 share the supercategory target and categories 5 to 28 share the supercategory clutter, so a detector can be trained on the two supercategories or on every kind. The ids are fixed: a kind that never appears in a swath keeps its id. Every clutter annotation also keeps its kind in attributes.shape and its origin, rock_field or placed_object, in attributes.source. Seafloor features that are not objects, such as ripple zones, bathymetry operators, trawl scars, pockmarks and outcrops, are recorded with their parameters in the image record rather than drawn as boxes. Every target and clutter object is placed at least 1.5 m inside the image, so no label is clipped by an edge.

The table counts labeled instances by category and split.

class	train	val	test	total
Targets
`cylinder`	402	51	47	500
`truncated_cone`	417	48	35	500
`wedge_belly`	404	48	48	500
`wedge_stl`	425	45	30	500
targets, 4 classes	1,648	192	160	2,000
Clutter
`rock`	70,526	5,881	10,223	86,630
`cross`	133	14	15	162
`torus`	125	11	17	153
`cone`	127	19	6	152
`dome`	123	10	9	142
`polyhedron`	120	10	10	140
`pyramid`	112	13	14	139
`hexprism`	112	13	13	138
`capsule`	111	14	8	133
`frustum`	109	10	13	132
`mound`	105	13	14	132
`drum`	99	13	16	128
`slab`	95	12	16	123
`fish_trap_square`	69	7	5	81
`ladder`	56	16	8	80
`tire`	64	7	5	76
`ship_propeller`	63	1	10	74
`keg`	50	10	12	72
`europalette`	58	8	5	71
`debris1`	59	5	4	68
`sunken_anchor`	54	6	8	68
`barrel_oil`	48	8	10	66
`rock_pile`	55	5	4	64
`clutter_other`	not used	not used	not used	not used
clutter, 23 classes	72,473	6,106	10,445	89,024
all labels, 27 classes	74,121	6,298	10,605	91,024

Files

Files per swath, shown for three seafloors

The figures below are drawn from the published files of three swaths simulated at 10 000 scatterers per square meter; the scene name and seafloor family are printed in the corner of each figure. The three scenes were selected so that together they contain every label category. All files share one orientation: along track up, range to the right, and the first row of each file at the top.

*_tdbp_drc.tif: The image is an 8000 by 3200 px, 8-bit palette GeoTIFF with the 256-entry ApertureLab sonar palette and deflate compression. A reader that ignores the palette gets the gray display image.
*_tdbp_coco.json: The file holds the labels in the 27 categories. Its image record holds the acquisition (altitude, sound speed, noise, residual motion), the tone map and the scene's features, and lists the sibling products.
*_dem.png: The file stores the composed seafloor, after bathymetry, roughness, ripples, objects and scour, as a 16-bit PNG on 5 cm cells. Height in meters is v / 65535 × (z_max_m − z_min_m) + z_min_m, with both constants in the file's own metadata.
*_rocks.csv: The table has one row per simulated boulder, giving its group, position, bed height, diameter, target strength and burial.
scenes/*.yaml: The scene file that produced the swath can be edited and run again in ApertureLab.

Rippled sand

A 40 m by 20 m window of the swath at full resolution with its label polygons drawn — **A window, labeled**40 m by 20 m at full resolution; target classes in green with their names, clutter objects in magenta, boulders in orange, shadow polygons dashed

Shaded height map of the rippled seafloor with a color bar in meters — **The height map**the 16-bit DEM decoded to meters with the scale in its own metadata, cropped to the imaged 80 m and shaded

The swath in gray with each boulder of its rocks table marked by diameter — **The boulder table**each row of the rocks CSV in the imaged band, placed with the beamformer's mapping and colored and sized by diameter

A rocky seafloor

A 40 m by 20 m window of the rocky swath with target, clutter and boulder polygons drawn — **A window, labeled**40 m by 20 m at full resolution, with the same colors as above; boulder labels are drawn in orange

Shaded height map of the rocky seafloor with a color bar in meters — **The height map**the DEM over the imaged band, height about the scene datum in meters

The rocky swath in gray with its boulders marked by diameter — **The boulder table**the rocks CSV in the imaged band; only boulders of 0.75 m and larger become clutter annotations

A sediment patchwork

Whole v1.1 swath of a patchwork of sediment types — **The image**the whole swath, reduced from 8000 by 3200 px

A 40 m by 20 m window of the patchwork swath with its label polygons drawn — **A window, labeled**40 m by 20 m at full resolution, with the same colors as above

Shaded height map of the patchwork seafloor with a color bar in meters — **The height map**bathymetry, tilt and bedforms together, height about the scene datum in meters

The patchwork swath in gray with its boulders marked by diameter — **The boulder table**the rocks CSV in the imaged band, colored and sized by diameter

Georeference

Coordinate frame

Each image is georeferenced so that GIS software displays it at a consistent scale and orientation. The frame is a transverse Mercator projection centered on the intersection of the equator and the prime meridian, on the WGS 84 datum, with unit scale factor and no false origin:

+proj=tmerc +lat_0=0 +lon_0=0 +k=1 +x_0=0 +y_0=0 +datum=WGS84 +units=m

The swaths tile edge to edge northward from (0, 0): swath si_NNNN spans northings NNNN × 80 m to (NNNN + 1) × 80 m. The release therefore forms a single strip 80 km long and 200 m wide. Northing runs along track, and easting is ground range to starboard in meters. The location is a convention for GIS software and does not correspond to a place on the seafloor. The metadata in each image and in its COCO file retain the scene's own along-track coordinates.

Scope

Limitations

The data are simulated; the simulator is checked against published scattering models and the theoretical point-spread function (see Fidelity), but no image is field data.
All swaths come from one sonar on a straight track, imaging the starboard side only.
The receiver-noise value in the manifest is a nominal ratio set against one reference seafloor; the signal-to-noise ratio measured in a given image also depends on the backscatter of that scene.
Radiometry is normalized by the cross-ping gain control instead of the aperture, so the mean brightness of a sand scene rises by a factor of about 2.8 from 30 m to 190 m of range.

Back to all datasets, or read how the labels are emitted.