Models

The lab maintains a collection of scientific models and methods that we can run on our own compute. You can bring us data, a research question, or both.

Each model runs in a version-controlled container, so we can preserve the computational environment used for an analysis. You can browse the catalog by research area, data type, or license. Models not listed here can usually be added on request. For larger workflows that combine several tools, see Process.

Unrestricted license only

Showing 64 of 64

Model	What it does	Field	Data	License
3DFinClassical method	Measures stem diameter, height, and taper from terrestrial scans without machine learning	Forestry & vegetation	point-cloud	GPL-3.0restricted use
ADAFDeep learning	Detects archaeological features such as barrows, ringforts and enclosures in lidar terrain models.	Geomorphology	imagery	Apache-2.0 code, CC-BY-SA-4.0 weightsrestricted use
AFwizardClassical method	Filters ground points from lidar with a different filter applied to each terrain segment.	Geomorphology	point-cloud	MIT
AIFSDeep learning	Produces global medium-range weather forecasts at about 31 km resolution, six-hourly.	Weather & climate	gridded-fields	Apache-2.0
AlphaFold 3Deep learning	Predicts the structure of proteins and their complexes with ligands and nucleic acids	Biology & genomics	sequence	CC-BY-NC-SA-4.0restricted use
AMS3D crown segmentationClassical method	Delineates tree crowns in drone lidar using adaptive mean-shift clustering.	Forestry & vegetation	point-cloud	GPL-3.0-or-laterrestricted use
autoXRDDeep learning	Identifies mineral phases in powder X-ray diffraction patterns	Materials & mineralogy	spectra	MIT
BiRefNetDeep learning	Produces one high-resolution foreground mask per image, holding edges on thin structures.	Cross-domain	imagery	MIT
Boltz-2Deep learning	Predicts structures of molecular complexes and estimates binding affinity.	Biology & genomics	sequence	MIT
Chai-1Deep learning	Predicts structures of proteins and their complexes.	Biology & genomics	sequence	Apache-2.0
ClayFoundation model	Creates embeddings from satellite imagery that can be used for classification and other downstream tasks with relatively few labeled examples.	Remote sensing	imagery	Apache-2.0
CLEANDeep learning	Predicts enzyme function (EC numbers) from protein sequence	Biology & genomics	sequence	Research use onlyrestricted use
ConcertoFoundation model	Creates point-cloud embeddings learned jointly from 3D scans and 2D imagery.	Cross-domain	point-cloud, imagery	Apache-2.0 code, CC-BY-NC-4.0 weightsrestricted use
CorrDiffDeep learning	Downscales ERA5 reanalysis to kilometre-scale fields over Europe.	Weather & climate	gridded-fields	Apache-2.0
Crater DetectionDeep learning	Detects lunar craters and matches them against a catalog to fix absolute position.	Planetary science	imagery	MIT
CROMAFoundation model	Creates embeddings from paired radar and optical satellite imagery.	Remote sensing	imagery	MIT
CrossEarthFoundation model	Segments satellite imagery in regions and conditions it was not trained on.	Remote sensing	imagery	MIT
DBloopsClassical method	Measures grain size and maps boulders from 3D point clouds of rock surfaces.	Geomorphology	point-cloud	GPL-3.0restricted use
DeepForestDeep learning	Detects individual tree crowns in aerial RGB imagery	Forestry & vegetation	imagery	MIT
Depth Anything 3Foundation model	Recovers depth, multi-view geometry and camera pose from ordinary photographs.	Cross-domain	imagery	Apache-2.0
DINOv3Foundation model	Extracts dense visual features from ordinary photographs, used as a frozen backbone for downstream tasks.	Cross-domain	imagery	LicenseRef-DINOv3 (not open source)restricted use
DINOv3-SATFoundation model	Extracts dense visual features from aerial and satellite RGB imagery, used as a frozen backbone for downstream tasks.	Remote sensing	imagery	LicenseRef-DINOv3 (not open source)restricted use
DNABERT-SFoundation model	Creates embeddings from DNA sequences for comparing genomes, identifying functional regions, and other genomic analyses.	Biology & genomics	sequence	Apache-2.0
DOFAFoundation model	Handles imagery with any number of spectral bands using a single model, and returns embeddings.	Remote sensing	imagery	CC-BY-4.0
DOFA-CLIPFoundation model	Scores multispectral or RGB imagery against text descriptions, without task-specific training.	Remote sensing	imagery	CC-BY-NC-4.0restricted use
ESM CFoundation model	Creates protein-sequence embeddings for tasks such as function and property prediction.	Biology & genomics	sequence	MIT
Evo 2Foundation model	Models genomic sequences and can be used to score sequence variation and predict the effects of genetic variants.	Biology & genomics	sequence	Apache-2.0
ForestFormer3DDeep learning	Segments forest point clouds into individual trees and semantic classes in a single pass.	Forestry & vegetation	point-cloud	CC-BY-NC-4.0restricted use
FourCastNet 3Deep learning	Produces global weather forecasts at 0.25-degree resolution, generating an ensemble rather than a single trajectory.	Weather & climate	gridded-fields	Apache-2.0
FSCTDeep learning	Segments stems in terrestrial scans and fits cylinders to measure diameter.	Forestry & vegetation	point-cloud	GPL-3.0restricted use
GeoCLIPFoundation model	Estimates where in the world a photograph was taken.	Cross-domain	imagery	MIT
GeoLG-3DFaultNetDeep learning	Labels fault surfaces voxel by voxel within a 3D seismic reflection volume.	Geophysics	volume	MIT
Grounding DINOFoundation model	Detects objects named in a free-text prompt, with no fixed class list.	Cross-domain	imagery	Apache-2.0
hyliteClassical method	Maps minerals in hyperspectral scans of drillcore and outcrop.	Materials & mineralogy	imagery, spectra	MIT
InSAR phase unwrappingDeep learning	Converts wrapped InSAR interferogram phase into continuous line-of-sight ground displacement.	Geophysics	imagery	MIT code, CC-BY-4.0 weights
KPConvDeep learning	Segments 3D point clouds using convolution defined directly on points rather than on a grid.	Cross-domain	point-cloud	MIT
MOMOFoundation model	Analyzes Mars orbital imagery for craters, boulders, and surface landforms	Planetary science	imagery	MIT AND CC-BY-4.0
Myria3DDeep learning	Segments airborne lidar into ground, vegetation, building, water, bridge and permanent structure.	Remote sensing	point-cloud	BSD-3-Clause
N2N4MDeep learning	Removes noise from Mars CRISM shortwave-infrared hyperspectral data.	Planetary science	imagery, spectra	MIT
NeuralHydrologyDeep learning	Predicts streamflow from rainfall using LSTM models	Hydrology	time-series	BSD-3-Clause
NTv3Foundation model	Produces genomic representations and predicts thousands of functional tracks across long stretches of sequence.	Biology & genomics	sequence	Non-commercial, gatedrestricted use
OctFormerDeep learning	Segments large 3D point clouds using an octree to keep attention tractable.	Cross-domain	point-cloud	MIT
Point Transformer V3Deep learning	Assigns a semantic class to every point in a 3D scan.	Cross-domain	point-cloud	MIT
PointsToWoodDeep learning	Separates leaf from wood in high-resolution terrestrial lidar	Forestry & vegetation	point-cloud	AGPL-3.0restricted use
Prithvi-EOFoundation model	Earth-observation foundation model developed by NASA and IBM, with models adapted for tasks including burn-scar, flood, and crop mapping.	Remote sensing	imagery	Apache-2.0
Raman mineral classifierClassical method	Identifies minerals from Raman spectra by matching against the RRUFF database	Materials & mineralogy	spectra	MIT
RemoteCLIPFoundation model	Links satellite imagery with natural-language descriptions, allowing images to be searched or classified without task-specific training.	Remote sensing	imagery	Apache-2.0
RF-DETRDeep learning	Detects objects in real time, and fine-tunes on a small set of local annotations.	Cross-domain	imagery	Apache-2.0
SAM 2Foundation model	Segments objects in images and video without task-specific training.	Cross-domain	imagery	Apache-2.0 AND BSD-3-Clause
SaProtFoundation model	Creates protein embeddings that use structure as well as sequence.	Biology & genomics	sequence, structure	MIT AND GPL-3.0restricted use
SatlasFoundation model	Pretrained backbones for satellite and aerial imagery, used as a starting point for task-specific models.	Remote sensing	imagery	Apache-2.0
SegmentAnyTreeDeep learning	Separates a forest lidar point cloud into individual trees	Forestry & vegetation	point-cloud	MIT
SeisBenchDeep learning	Picks earthquake phase arrivals from seismic waveforms	Geophysics	waveform	GPL-3.0restricted use
SeisTDeep learning	Extracts several earthquake measurements from a single waveform: phase arrivals, polarity, magnitude, back-azimuth, and distance.	Geophysics	waveform	MIT
SigLIP 2Foundation model	Scores images against text, for zero-shot triage and text search across image collections.	Cross-domain	imagery	Apache-2.0
SonataFoundation model	Creates embeddings from 3D point clouds without requiring labelled training data.	Cross-domain	point-cloud	Apache-2.0 code, CC-BY-NC-4.0 weightsrestricted use
StormCastDeep learning	Nowcasts convective storms over the continental United States on a 3 km grid.	Weather & climate	gridded-fields	Apache-2.0
SuperPoint TransformerDeep learning	Performs semantic and panoptic segmentation of very large scans by grouping points before classifying them.	Cross-domain	point-cloud	MIT
TerraMindFoundation model	Multimodal Earth-observation foundation model that combines radar, optical, elevation, and land-cover data, and can generate one modality from another.	Remote sensing	imagery	Apache-2.0
TimesFMFoundation model	Forecasts regularly sampled time series without training a new model for each dataset.	Cross-domain	time-series	Apache-2.0
TreeLearnDeep learning	Segments individual trees from ground-based and mobile lidar scans	Forestry & vegetation	point-cloud	MIT
treeXClassical method	Separates individual trees from laser scans without any trained model.	Forestry & vegetation	point-cloud	MIT
TRELLIS.2Foundation model	Generates a textured 3D mesh with materials from a single image.	Cross-domain	imagery	MIT
UtoniaFoundation model	Creates embeddings from remote-sensing point clouds without requiring labelled training data.	Cross-domain	point-cloud	Apache-2.0 code, CC-BY-NC-4.0 weightsrestricted use

3DFin
Measures stem diameter, height, and taper from terrestrial scans without machine learning
Forestry & vegetation · Classical method · point-cloud · GPL-3.0 (restricted use)
ADAF
Detects archaeological features such as barrows, ringforts and enclosures in lidar terrain models.
Geomorphology · Deep learning · imagery · Apache-2.0 code, CC-BY-SA-4.0 weights (restricted use)
AFwizard
Filters ground points from lidar with a different filter applied to each terrain segment.
Geomorphology · Classical method · point-cloud · MIT
AIFS
Produces global medium-range weather forecasts at about 31 km resolution, six-hourly.
Weather & climate · Deep learning · gridded-fields · Apache-2.0
AlphaFold 3
Predicts the structure of proteins and their complexes with ligands and nucleic acids
Biology & genomics · Deep learning · sequence · CC-BY-NC-SA-4.0 (restricted use)
AMS3D crown segmentation
Delineates tree crowns in drone lidar using adaptive mean-shift clustering.
Forestry & vegetation · Classical method · point-cloud · GPL-3.0-or-later (restricted use)
autoXRD
Identifies mineral phases in powder X-ray diffraction patterns
Materials & mineralogy · Deep learning · spectra · MIT
BiRefNet
Produces one high-resolution foreground mask per image, holding edges on thin structures.
Cross-domain · Deep learning · imagery · MIT
Boltz-2
Predicts structures of molecular complexes and estimates binding affinity.
Biology & genomics · Deep learning · sequence · MIT
Chai-1
Predicts structures of proteins and their complexes.
Biology & genomics · Deep learning · sequence · Apache-2.0
Clay
Creates embeddings from satellite imagery that can be used for classification and other downstream tasks with relatively few labeled examples.
Remote sensing · Foundation model · imagery · Apache-2.0
CLEAN
Predicts enzyme function (EC numbers) from protein sequence
Biology & genomics · Deep learning · sequence · Research use only (restricted use)
Concerto
Creates point-cloud embeddings learned jointly from 3D scans and 2D imagery.
Cross-domain · Foundation model · point-cloud, imagery · Apache-2.0 code, CC-BY-NC-4.0 weights (restricted use)
CorrDiff
Downscales ERA5 reanalysis to kilometre-scale fields over Europe.
Weather & climate · Deep learning · gridded-fields · Apache-2.0
Crater Detection
Detects lunar craters and matches them against a catalog to fix absolute position.
Planetary science · Deep learning · imagery · MIT
CROMA
Creates embeddings from paired radar and optical satellite imagery.
Remote sensing · Foundation model · imagery · MIT
CrossEarth
Segments satellite imagery in regions and conditions it was not trained on.
Remote sensing · Foundation model · imagery · MIT
DBloops
Measures grain size and maps boulders from 3D point clouds of rock surfaces.
Geomorphology · Classical method · point-cloud · GPL-3.0 (restricted use)
DeepForest
Detects individual tree crowns in aerial RGB imagery
Forestry & vegetation · Deep learning · imagery · MIT
Depth Anything 3
Recovers depth, multi-view geometry and camera pose from ordinary photographs.
Cross-domain · Foundation model · imagery · Apache-2.0
DINOv3
Extracts dense visual features from ordinary photographs, used as a frozen backbone for downstream tasks.
Cross-domain · Foundation model · imagery · LicenseRef-DINOv3 (not open source) (restricted use)
DINOv3-SAT
Extracts dense visual features from aerial and satellite RGB imagery, used as a frozen backbone for downstream tasks.
Remote sensing · Foundation model · imagery · LicenseRef-DINOv3 (not open source) (restricted use)
DNABERT-S
Creates embeddings from DNA sequences for comparing genomes, identifying functional regions, and other genomic analyses.
Biology & genomics · Foundation model · sequence · Apache-2.0
DOFA
Handles imagery with any number of spectral bands using a single model, and returns embeddings.
Remote sensing · Foundation model · imagery · CC-BY-4.0
DOFA-CLIP
Scores multispectral or RGB imagery against text descriptions, without task-specific training.
Remote sensing · Foundation model · imagery · CC-BY-NC-4.0 (restricted use)
ESM C
Creates protein-sequence embeddings for tasks such as function and property prediction.
Biology & genomics · Foundation model · sequence · MIT
Evo 2
Models genomic sequences and can be used to score sequence variation and predict the effects of genetic variants.
Biology & genomics · Foundation model · sequence · Apache-2.0
ForestFormer3D
Segments forest point clouds into individual trees and semantic classes in a single pass.
Forestry & vegetation · Deep learning · point-cloud · CC-BY-NC-4.0 (restricted use)
FourCastNet 3
Produces global weather forecasts at 0.25-degree resolution, generating an ensemble rather than a single trajectory.
Weather & climate · Deep learning · gridded-fields · Apache-2.0
FSCT
Segments stems in terrestrial scans and fits cylinders to measure diameter.
Forestry & vegetation · Deep learning · point-cloud · GPL-3.0 (restricted use)
GeoCLIP
Estimates where in the world a photograph was taken.
Cross-domain · Foundation model · imagery · MIT
GeoLG-3DFaultNet
Labels fault surfaces voxel by voxel within a 3D seismic reflection volume.
Geophysics · Deep learning · volume · MIT
Grounding DINO
Detects objects named in a free-text prompt, with no fixed class list.
Cross-domain · Foundation model · imagery · Apache-2.0
hylite
Maps minerals in hyperspectral scans of drillcore and outcrop.
Materials & mineralogy · Classical method · imagery, spectra · MIT
InSAR phase unwrapping
Converts wrapped InSAR interferogram phase into continuous line-of-sight ground displacement.
Geophysics · Deep learning · imagery · MIT code, CC-BY-4.0 weights
KPConv
Segments 3D point clouds using convolution defined directly on points rather than on a grid.
Cross-domain · Deep learning · point-cloud · MIT
MOMO
Analyzes Mars orbital imagery for craters, boulders, and surface landforms
Planetary science · Foundation model · imagery · MIT AND CC-BY-4.0
Myria3D
Segments airborne lidar into ground, vegetation, building, water, bridge and permanent structure.
Remote sensing · Deep learning · point-cloud · BSD-3-Clause
N2N4M
Removes noise from Mars CRISM shortwave-infrared hyperspectral data.
Planetary science · Deep learning · imagery, spectra · MIT
NeuralHydrology
Predicts streamflow from rainfall using LSTM models
Hydrology · Deep learning · time-series · BSD-3-Clause
NTv3
Produces genomic representations and predicts thousands of functional tracks across long stretches of sequence.
Biology & genomics · Foundation model · sequence · Non-commercial, gated (restricted use)
OctFormer
Segments large 3D point clouds using an octree to keep attention tractable.
Cross-domain · Deep learning · point-cloud · MIT
Point Transformer V3
Assigns a semantic class to every point in a 3D scan.
Cross-domain · Deep learning · point-cloud · MIT
PointsToWood
Separates leaf from wood in high-resolution terrestrial lidar
Forestry & vegetation · Deep learning · point-cloud · AGPL-3.0 (restricted use)
Prithvi-EO
Earth-observation foundation model developed by NASA and IBM, with models adapted for tasks including burn-scar, flood, and crop mapping.
Remote sensing · Foundation model · imagery · Apache-2.0
Raman mineral classifier
Identifies minerals from Raman spectra by matching against the RRUFF database
Materials & mineralogy · Classical method · spectra · MIT
RemoteCLIP
Links satellite imagery with natural-language descriptions, allowing images to be searched or classified without task-specific training.
Remote sensing · Foundation model · imagery · Apache-2.0
RF-DETR
Detects objects in real time, and fine-tunes on a small set of local annotations.
Cross-domain · Deep learning · imagery · Apache-2.0
SAM 2
Segments objects in images and video without task-specific training.
Cross-domain · Foundation model · imagery · Apache-2.0 AND BSD-3-Clause
SaProt
Creates protein embeddings that use structure as well as sequence.
Biology & genomics · Foundation model · sequence, structure · MIT AND GPL-3.0 (restricted use)
Satlas
Pretrained backbones for satellite and aerial imagery, used as a starting point for task-specific models.
Remote sensing · Foundation model · imagery · Apache-2.0
SegmentAnyTree
Separates a forest lidar point cloud into individual trees
Forestry & vegetation · Deep learning · point-cloud · MIT
SeisBench
Picks earthquake phase arrivals from seismic waveforms
Geophysics · Deep learning · waveform · GPL-3.0 (restricted use)
SeisT
Extracts several earthquake measurements from a single waveform: phase arrivals, polarity, magnitude, back-azimuth, and distance.
Geophysics · Deep learning · waveform · MIT
SigLIP 2
Scores images against text, for zero-shot triage and text search across image collections.
Cross-domain · Foundation model · imagery · Apache-2.0
Sonata
Creates embeddings from 3D point clouds without requiring labelled training data.
Cross-domain · Foundation model · point-cloud · Apache-2.0 code, CC-BY-NC-4.0 weights (restricted use)
StormCast
Nowcasts convective storms over the continental United States on a 3 km grid.
Weather & climate · Deep learning · gridded-fields · Apache-2.0
SuperPoint Transformer
Performs semantic and panoptic segmentation of very large scans by grouping points before classifying them.
Cross-domain · Deep learning · point-cloud · MIT
TerraMind
Multimodal Earth-observation foundation model that combines radar, optical, elevation, and land-cover data, and can generate one modality from another.
Remote sensing · Foundation model · imagery · Apache-2.0
TimesFM
Forecasts regularly sampled time series without training a new model for each dataset.
Cross-domain · Foundation model · time-series · Apache-2.0
TreeLearn
Segments individual trees from ground-based and mobile lidar scans
Forestry & vegetation · Deep learning · point-cloud · MIT
treeX
Separates individual trees from laser scans without any trained model.
Forestry & vegetation · Classical method · point-cloud · MIT
TRELLIS.2
Generates a textured 3D mesh with materials from a single image.
Cross-domain · Foundation model · imagery · MIT
Utonia
Creates embeddings from remote-sensing point clouds without requiring labelled training data.
Cross-domain · Foundation model · point-cloud · Apache-2.0 code, CC-BY-NC-4.0 weights (restricted use)

This is what we currently have set up, not a fixed list. If there is a model or method you want to use, contact us. In many cases, we can add it to the lab environment.

Contact the Lab