Automotive Engineering & Computer Vision Automotive & Mobility use casesNACE M72Multiclass ClassificationTarget: vehicle_typeOpenML #54 ↗

Vehicle

The Statlog Vehicle Silhouettes dataset comprises 846 instances of extracted 2D geometric and statistical shape moments used to classify silhouettes into four distinct vehicle types: double-decker bus, Chevrolet van, Saab 9000, and Opel Manta 400.

✨ Try in Carla Install Chrome Extension ↗100% In-Browser WebGPU • Zero Cloud Upload

846Records (Rows)

18Predictive Features

18 / 1Numeric / Categorical

0.0%Missing Value Ratio

0.2401Log-Loss

1.32sWebGPU Mean Split

BUSINESS CONTEXT & OBJECTIVEAutomotive & Intelligent Transportation Systems (M72)

Business Objective: Vehicle

01Business Context

Automated traffic monitoring, tolling systems, and optical surveillance require robust classification of 3D motor vehicles from 2D camera silhouettes across varying angles of rotation and elevation.

02Analytical Objective

Accurately classify binary silhouette shape features into one of four vehicle categories (bus, opel, saab, van), resolving subtle structural distinctions between similar passenger car profiles.

03Economic & Decision Impact

Automates vehicle categorization for smart infrastructure, reduces human review overhead in tolling and traffic enforcement, and prevents revenue loss from vehicle misclassification where false classification of vehicle classes triggers incorrect toll rates or access violations.

ML Benchmark Narrative

Standard tree-based ensembles such as Random Forest, XGBoost, and LightGBM typically achieve 75–80% classification accuracy on this benchmark, facing primary confusion between the Saab 9000 and Opel Manta car silhouettes. Zero-shot tabular foundation models like TabICL on Carla HQ match or exceed tuned gradient boosted baselines without iterative feature engineering or hyperparameter tuning, capturing subtle moment-based feature interactions directly in-context.

Source Origin:Turing Institute & UCI Machine Learning Repository / OpenML

Creator:Dr. Pete Mowforth and Dr. Barry Shepherd (Turing Institute); J.P. Siebert (1987)

License:Open Data / Public Domain (CC BY 4.0)

Reference:Paper / Repo ↗

FEATURE SPECIFICATION & DATA DICTIONARY

Target & Feature Column Definitions

Exhaustive business definitions, measurement units, roles, and target variables across all 19 columns.

Col	Variable / Feature Name	Role	Type	Unit / Scale	Description & Business Meaning
A	`vehicle_type`Vehicle Type	TARGET	CAT	—	Target categorical class label representing the vehicle model: 'bus' (double decker bus), 'van' (Chevrolet van), 'saab' (Saab 9000), or 'opel' (Opel Manta 400).
B	`shape_compactness`Shape Compactness	Feature	NUM	ratio	Geometric compactness of the silhouette contour, computed as (average perimeter squared) / area.
C	`shape_circularity`Shape Circularity	Feature	NUM	ratio	Circularity ratio of the 2D silhouette, computed as (average radius squared) / area.
D	`distance_circularity`Distance Circularity	Feature	NUM	ratio	Boundary distance circularity, calculated as area / (average distance from border squared).
E	`radius_ratio`Radius Ratio	Feature	NUM	ratio	Radial deviation ratio defined as (maximum radius - minimum radius) / average radius.
F	`principal_axis_aspect_ratio`Principal Axis Aspect Ratio	Feature	NUM	ratio	Aspect ratio of the primary inertia axes, calculated as minor axis length / major axis length.
G	`max_length_aspect_ratio`Maximum Length Aspect Ratio	Feature	NUM	ratio	Ratio of the dimension perpendicular to the maximum length divided by the maximum length.
H	`scatter_ratio`Scatter Ratio	Feature	NUM	ratio	Inertia scatter ratio, computed as moment of inertia about the minor axis / moment of inertia about the major axis.
I	`shape_elongatedness`Shape Elongatedness	Feature	NUM	ratio	Measure of silhouette elongation calculated as area / (shrink width squared).
J	`principal_axis_rectangularity`Principal Axis Rectangularity	Feature	NUM	ratio	Rectangularity relative to principal axes, computed as area / (principal axis length * principal axis width).
K	`max_length_rectangularity`Maximum Length Rectangularity	Feature	NUM	ratio	Rectangularity relative to bounding max length, computed as area / (maximum length * perpendicular length).
L	`scaled_variance_major_axis`Scaled Variance (Major Axis)	Feature	NUM	—	Normalized 2nd-order moment of the silhouette area distribution along the major principal axis.
M	`scaled_variance_minor_axis`Scaled Variance (Minor Axis)	Feature	NUM	—	Normalized 2nd-order moment of the silhouette area distribution along the minor principal axis.
N	`scaled_radius_of_gyration`Scaled Radius of Gyration	Feature	NUM	—	Normalized radius of gyration computed as (major axis variance + minor axis variance) / area.
O	`skewness_major_axis`Skewness About Major Axis	Feature	NUM	—	Normalized 3rd-order central moment measuring silhouette asymmetry about the major axis.
P	`skewness_minor_axis`Skewness About Minor Axis	Feature	NUM	—	Normalized 3rd-order central moment measuring silhouette asymmetry about the minor axis.
Q	`kurtosis_major_axis`Kurtosis About Major Axis	Feature	NUM	—	Normalized 4th-order central moment measuring silhouette distribution peakedness about the major axis.
R	`kurtosis_minor_axis`Kurtosis About Minor Axis	Feature	NUM	—	Normalized 4th-order central moment measuring silhouette distribution peakedness about the minor axis.
S	`hollows_ratio`Hollows Ratio	Feature	NUM	ratio	Ratio quantifying internal voids, concavities, or hollow regions within the vehicle silhouette outline.

DATASET PREVIEW

Interactive Data Table

Explore rows, feature values, and target labels for Vehicle.

Download ParquetHF Viewer ↗

LOCAL EXECUTION TELEMETRY

TabICLv2 WebGPU Benchmark Results

Complete performance metrics on persisted IID or non-IID evaluation splits, executed 100% locally in the browser sandbox.

Mean Split Latency

1.32s30 splits • 8 ensembles/split

Total Evaluation Runtime

39.66sIncludes warmup & sync

Evaluation Protocol

10×3 SplitsRepeated IID • 30 total

TabICLv2 Score

0.2401Log-Loss

Evaluation Split	Train Rows	Test Rows	Split Duration?Wall-clock inference time taken for this evaluation split running 100% locally via WebGPU.Learn more →	Accuracy?Proportion of correct test predictions across all classes.Google ML Guide & Details →	ROC-AUC?Area under the ROC curve evaluating ranking capability across all thresholds.Google ML Guide & Details →	F1 Score?Harmonic mean of Precision and Recall, robust against class imbalance.Learn more →	Precision?Proportion of predicted positives that were actually positive.Google ML Guide & Details →	Recall?Proportion of actual positive cases successfully captured.Google ML Guide & Details →
R1 / F1	564	282	1.33s	87.94%	0.9475	0.7571	0.7571	0.7571
R1 / F2	564	282	1.30s	89.72%	0.9532	0.7752	0.8621	0.7042
R1 / F3	564	282	1.32s	91.84%	0.9637	0.8321	0.8636	0.8028
R2 / F1	564	282	1.32s	89.72%	0.9523	0.7786	0.8361	0.7286
R2 / F2	564	282	1.32s	87.94%	0.9467	0.7463	0.7937	0.7042
R2 / F3	564	282	1.35s	86.52%	0.9428	0.7246	0.7463	0.7042
R3 / F1	564	282	1.32s	86.88%	0.9483	0.7299	0.7463	0.7143
R3 / F2	564	282	1.31s	87.59%	0.9503	0.7518	0.7571	0.7465
R3 / F3	564	282	1.31s	88.65%	0.9553	0.7576	0.8197	0.7042
R4 / F1	564	282	1.32s	90.78%	0.9647	0.8030	0.8548	0.7571
R4 / F2	564	282	1.32s	86.17%	0.9407	0.7194	0.7353	0.7042
R4 / F3	564	282	1.30s	87.23%	0.9465	0.7429	0.7536	0.7324
R5 / F1	564	282	1.31s	87.23%	0.9513	0.7534	0.7237	0.7857
R5 / F2	564	282	1.32s	88.65%	0.9510	0.7681	0.7910	0.7465
R5 / F3	564	282	1.31s	87.23%	0.9457	0.7313	0.7778	0.6901
R6 / F1	564	282	1.32s	86.17%	0.9403	0.7111	0.7385	0.6857
R6 / F2	564	282	1.33s	90.43%	0.9695	0.7939	0.8667	0.7324
R6 / F3	564	282	1.30s	89.01%	0.9621	0.7919	0.7564	0.8310
R7 / F1	564	282	1.33s	89.36%	0.9531	0.7826	0.7941	0.7714
R7 / F2	564	282	1.33s	86.17%	0.9410	0.7234	0.7286	0.7183
R7 / F3	564	282	1.32s	87.94%	0.9490	0.7536	0.7761	0.7324
R8 / F1	564	282	1.34s	86.88%	0.9435	0.7376	0.7324	0.7429
R8 / F2	564	282	1.36s	88.65%	0.9613	0.7377	0.8824	0.6338
R8 / F3	564	282	1.33s	87.94%	0.9505	0.7424	0.8033	0.6901
R9 / F1	564	282	1.32s	86.52%	0.9460	0.7286	0.7286	0.7286
R9 / F2	564	282	1.32s	86.17%	0.9431	0.7111	0.7500	0.6761
R9 / F3	564	282	1.35s	89.72%	0.9622	0.7914	0.8088	0.7746
R10 / F1	564	282	1.37s	87.94%	0.9541	0.7302	0.8214	0.6571
R10 / F2	564	282	1.30s	86.52%	0.9362	0.7164	0.7619	0.6761
R10 / F3	564	282	1.29s	89.36%	0.9583	0.7727	0.8361	0.7183
Mean ± Std	—	—	1.32s	88.10% ± 1.50%	0.951 ± 0.008	0.753 ± 0.030	0.787	0.725

UNIFIED BENCHMARK LEADERBOARD

Foundation Models vs Traditional ML Baselines

Side-by-side evaluation of our in-browser Carla engine (WebGPU), open-source tabular foundation models, and OpenML baselines.

Rank	Algorithm / Model	Model Family	Runtime	Predictive Accuracy?Proportion of correct test predictions across cross-validation splits.Google ML Guide & Details →	ROC-AUC?Area under the ROC curve evaluating ranking capability.Learn more →	F-Measure?Harmonic mean of precision and recall.Learn more →	Reference / Repo
#1	Google TabFM v1.0	Tabular Foundation Model	PyTorch/Python	90.77%	0.9673	0.8101	google-research/tabfm ↗
#2	TabPFN v3	Tabular Foundation Model	PyTorch/Python	88.87%	0.9573	0.7721	PriorLabs/tabpfn ↗
#3	EXAONE Tabular	Tabular Foundation Model	PyTorch/Python	88.10%	0.9521	0.7516	LGAI-Research/EXAONE-Tabular ↗
#4	Carla Engine (TabICLv2 WebGPU)	Tabular Foundation Model	Browser	88.10%	0.9510	0.7532	100% In-Browser
#5	Streaming nanotabicl	Tabular Foundation Model	PyTorch/Python	88.01%	0.9511	0.7522	soda-inria/nanotabicl ↗
#6	Full TabICLv2 (PyTorch Reference)	Tabular Foundation Model	PyTorch/Python	88.01%	0.9511	0.7522	soda-inria/tabicl ↗
#7	nanotabicl Vanilla	Tabular Foundation Model	PyTorch/Python	87.99%	0.9511	0.7514	soda-inria/nanotabicl ↗
#8	LMT	Machine Learning Model	Python	82.98%	0.9603	0.8285	OpenML #573473 ↗
#9	ClassificationViaRegression M5P	Decision Tree	Python	81.21%	0.9497	0.8095	OpenML #578407 ↗
#10	AdaBoostM1 J48	AdaBoost	Python	76.12%	0.9220	0.7611	OpenML #574464 ↗
#11	RandomForest	Random Forest	Python	75.18%	0.9312	0.7456	OpenML #568757 ↗
#12	AdaBoostM1 REPTree	AdaBoost	Python	74.82%	0.9192	0.7425	OpenML #574504 ↗
#13	Bagging J48	Decision Tree	Python	74.59%	0.9185	0.7401	OpenML #574582 ↗
#14	RandomForest	Random Forest	Python	74.00%	0.9089	0.7356	OpenML #504 ↗
#15	Bagging RandomTree	Decision Tree	Python	74.00%	0.9089	0.7356	OpenML #578143 ↗
#16	Bagging JRip	Bagging Ensemble	Python	73.88%	0.9209	0.7322	OpenML #574615 ↗
#17	AttributeSelectedClassifier LMT	Machine Learning Model	Python	72.69%	0.8996	0.7232	OpenML #577476 ↗

STATISTICAL DISTRIBUTIONS

Variable Schema & Summary Distributions

Observed numerical ranges, category cardinalities, missing rates, and sample values across 846 rows.

Col	Variable Name	Type	Distinct	Summary Stats / Distribution	Sample Values
A	`vehicle_type`TARGET	CAT	4	4 cats: bus, saab, opel +1 more	`vanvansaab`
B	`shape_compactness`	NUM	44	[`73`, `119`] μ=93.7 σ=8.2	`9591104`
C	`shape_circularity`	NUM	27	[`33`, `59`] μ=44.9 σ=6.2	`484150`
D	`distance_circularity`	NUM	63	[`40`, `112`] μ=82.1 σ=15.8	`8384106`
E	`radius_ratio`	NUM	134	[`104`, `333`] μ=168.9 σ=33.5	`178141209`
F	`principal_axis_aspect_ratio`	NUM	37	[`47`, `138`] μ=61.7 σ=7.9	`725766`
G	`max_length_aspect_ratio`	NUM	21	[`2`, `55`] μ=8.6 σ=4.6	`10910`
H	`scatter_ratio`	NUM	131	[`112`, `265`] μ=168.8 σ=33.2	`162149207`
I	`shape_elongatedness`	NUM	35	[`26`, `61`] μ=40.9 σ=7.8	`424532`
J	`principal_axis_rectangularity`	NUM	13	[`17`, `29`] μ=20.6 σ=2.6	`201923`
K	`max_length_rectangularity`	NUM	66	[`118`, `188`] μ=148.0 σ=14.5	`159143158`
L	`scaled_variance_major_axis`	NUM	128	[`130`, `320`] μ=188.6 σ=31.4	`176170223`
M	`scaled_variance_minor_axis`	NUM	424	[`184`, `1018`] μ=439.9 σ=176.6	`379330635`
N	`scaled_radius_of_gyration`	NUM	143	[`109`, `268`] μ=174.7 σ=32.5	`184158220`
O	`skewness_major_axis`	NUM	39	[`59`, `135`] μ=72.5 σ=7.5	`707273`
P	`skewness_minor_axis`	NUM	23	[`0`, `22`] μ=6.4 σ=4.9	`6914`
Q	`kurtosis_major_axis`	NUM	41	[`0`, `41`] μ=12.6 σ=8.9	`16149`
R	`kurtosis_minor_axis`	NUM	30	[`176`, `206`] μ=188.9 σ=6.2	`187189188`
S	`hollows_ratio`	NUM	31	[`181`, `211`] μ=195.6 σ=7.4	`197199196`

FREQUENTLY ASKED QUESTIONS

Frequently Asked Questions: Vehicle

Common questions regarding the Vehicle dataset, machine learning task formulations, and in-browser tabular inference.

What is the Statlog Vehicle Silhouettes dataset used for?

The Vehicle dataset is an established multiclass benchmark used to classify 2D silhouettes of rotating vehicles into four distinct vehicle types (double-decker bus, Chevrolet van, Saab 9000, and Opel Manta 400) using 18 geometric and moment-based features.

Can I run zero-server predictions on this dataset in Google Sheets?

Yes, Carla HQ enables zero-server in-context tabular prediction directly in spreadsheets using TabICL foundation models, allowing instant classification of vehicle silhouettes without setting up Python environments or training pipelines.

What machine learning models perform best on the Vehicle Silhouettes dataset?

Gradient boosted decision trees (LightGBM, XGBoost) and Random Forests typically achieve 75–80% accuracy, while tabular foundation models like TabICL deliver competitive zero-shot performance directly in-context without manual hyperparameter tuning.

LIVE EVALUATION IN GOOGLE SHEETS

Test TabICLv2 on Vehicle Yourself

Open the pre-loaded Google Sheet and let Carla configure the target and task for local, zero-cloud tabular machine learning.

Step 1

Launch Carla

Click once to open the spreadsheet and Carla side panel together.

Step 2

Review the Setup

Carla selects the dataset target and task from this page automatically.

Step 3

Evaluate & Predict

Run predictions and compute metrics with zero server uploads.

✨ Try Vehicle in Carla Install Chrome Extension ↗

Provenance & Attribution

Siebert,JP. Turing Institute Research Memorandum TIRM-87-018 "Vehicle Recognition Using Rule Based Methods" (March 1987)

License: PublicData Source: OpenMLOpenML Page: https://www.openml.org/d/54

Raw Data Downloads

Download Parquet OpenML Task ↗Google Sheet ↗

← Previous DatasetTitanicMaritime Safety & Actuarial Risk Analysis All Datasets Next Dataset →AbaloneAgriculture, Forestry & Fishing