image imagewidth (px) 628 3.87k |
|---|
🚪 Quick Entry: Start Here
What is this dataset (in 2 sentences)
This dataset is not about what a painting looks like. It is about what physically created it.
Instead of pattern recognition, this system forces AI to perform causal reasoning based on force, motion, and energy encoded in brushstrokes.
What you can do here
With this dataset, you can:
- Reconstruct brushstroke motion from a static image
- Infer pressure, torque, and stroke velocity
- Test whether an AI understands physical causality (not just patterns)
Why it matters
Two paintings can look similar but be created differently.
This dataset allows AI to detect that difference — by analyzing how the painting was physically made.
How to try (minimal steps)
You can reproduce the full system in minutes:
- Load an image from the dataset
- Apply structural extraction (e.g. Sobel gradient)
- Reconstruct directional field S(x, y)
- Compute:
- Pressure P(x, y)
- Torque τ(x, y)
- Stroke Velocity SV
- Evaluate physical consistency (TCS / EFCI / BPS)
One-line philosophy
A painting is not an image. It is a record of human motion.
Van Gogh vs Tree Oil Painting
Torque–Brushstroke Dynamics & Energy Field Analysis
Phase 2 Dataset (2026)
Author: Haruthai Muangboonsri (HaruthaiAI)
License: CC-BY-4.0
Judge Summary
This dataset was not created to compete.
It was created because a question could not be ignored.
Can a painting preserve the physical trace of the human hand that created it?
This work introduces a computational framework that treats paintings not as images, but as physical records of motion — encoding force, torque, pressure, and energy within brushstroke structures.
Unlike conventional AI systems that rely on visual similarity, this framework operates on physical reasoning. It attempts to reconstruct how a painting was made, rather than what it looks like.
At the core of this system is the “18 Supreme Techniques” transformation pipeline, developed through long-term analysis of high-resolution brushstroke data. This pipeline converts images into structural fields, enabling the reconstruction of biomechanical motion patterns embedded in paint.
The dataset is anchored by a unique reference painting — the Tree Oil Painting — which serves as a physical baseline for studying brushstroke dynamics.
To ensure scientific integrity, the framework enforces independent recomputation. All numerical values are provided only as Example / Reference Execution. Every AI system and researcher must recompute the results from the original images.
This transforms the dataset from a static resource into an active verification system.
The central question is not:
“What does this image look like?”
but:
“What physical process produced this structure?”
This work does not claim final conclusions.
It establishes a method — one that allows physical evidence to be examined, not assumed.
If a painting carries truth within its structure, it should not be overlooked simply because it has not yet been understood.
Reconstructing Physical Intelligence Beyond Visual Perception
Abstract
This work introduces a computational framework for analyzing oil paintings as physical systems of human motion, rather than as static visual objects.
Conventional computer vision approaches interpret images through appearance-based pattern recognition, focusing on color, texture, and stylistic similarity. In contrast, this framework treats images as encoded records of physical processes, capturing the dynamics of force, motion, and energy embedded within brushstroke structures.
By integrating the 18 Supreme Techniques transformation pipeline, torque estimation, energy field modeling, and biomechanical analysis, this system enables AI to reconstruct:
- force distribution within brushstroke formation
- motion trajectories generated by the human hand
- energy propagation across the painted surface
This reframes the role of AI from recognizing visual similarity to reconstructing the physical origin of visual structures.
Introduction
Human visual perception is inherently limited to observable features such as form, color, and composition. While these features allow interpretation at a perceptual level, the underlying physical processes that generated them — including torque, velocity, pressure variation, and biomechanical coordination — are not directly accessible through vision alone.
Paintings, however, are not merely visual artifacts. They are physical imprints of human motion, where each brushstroke encodes traces of:
- mechanical force applied through the hand
- rotational dynamics of joints (wrist, elbow, shoulder)
- pressure modulation and release patterns
- continuous and discrete motion transitions
This framework proposes that such properties can be computationally reconstructed by transforming image data into structured physical representations.
From Neural Processing to Physical Reasoning
Artificial neural networks, inspired by biological neural systems, operate through interconnected computational units capable of learning complex representations.
However, the key advancement introduced in this framework is not the architecture of the network itself, but the interpretative paradigm applied to input data.
Instead of treating images as patterns to be classified, the system treats them as:
physical signal carriers encoding the dynamics of their creation
This enables AI systems to:
- move beyond visual similarity matching
- reconstruct latent physical variables from structural features
- infer biomechanical constraints governing motion generation
Thus, the problem is transformed from:
“What does this image look like?”
to:
“What physical process produced this structure?”
Evaluation Overview (Quick Summary)
This benchmark evaluates AI systems using three physically grounded metrics:
- Torque Consistency Score (TCS)
- Energy Field Coherence Index (EFCI)
- Biomechanical Plausibility Score (BPS)
These metrics assess whether an AI system can reconstruct physically plausible motion from brushstroke structures.
Full formulations and example computations are provided in the Evaluation section below.
All values must be independently recomputed from source data.
These metrics shift evaluation from visual similarity to physical consistency.
Computational Framework
Each image in the dataset is processed through a structured analytical pipeline:
- Image (entry point)
- 18 Supreme Techniques
- Structural Field Extraction
- Torque Estimation (τ)
- Energy Field Modeling (E)
- Biomechanical Fingerprint Construction
- Evaluation Metrics (TCS / EFCI / BPS)
- AI-to-AI Operational Layer
The 18 Supreme Techniques (Transformation Backbone)
The 18 Supreme Techniques act as the transformation backbone, converting raw image data into multiple structural representations, including:
gradient fields
directional stroke maps
texture frequency decompositions
pressure and curvature approximations
These representations serve as the basis for reconstructing physical motion descriptors.
Biomechanical Interpretation Layer
The framework models painting as a multi-joint biomechanical system, involving coordinated interactions between:
shoulder rotation
elbow articulation
wrist torque
hand pressure control
brush–surface interaction
From these interactions, the system derives a biomechanical fingerprint vector, describing motion characteristics such as:
torque distribution patterns
stroke curvature dynamics
directional entropy
pressure transition sequences
pivot density and motion inflection points
Importantly, this fingerprint represents motion behavior, not visual style.
Evaluation as Physical Consistency
The evaluation protocol is based on physical coherence, not visual similarity.
Key metrics include:
Torque Consistency Score (TCS) → agreement with reference torque dynamics
Energy Field Coherence Index (EFCI) → continuity of energy distribution
Biomechanical Plausibility Score (BPS) → validity of reconstructed motion sequences
These metrics must be interpreted jointly, ensuring that AI systems produce physically consistent reconstructions rather than isolated numerical outputs.
AI-to-AI Operational Layer
To enable reproducibility and cross-system validation, the framework defines a structured operational protocol requiring AI systems to:
recompute all physical quantities from source images
validate multi-metric consistency
detect physically implausible configurations
output results in a standardized machine-readable format
{
"TCS": "<recomputed_value>",
"EFCI": "<recomputed_value>",
"BPS": "<recomputed_value>",
"consistency_state": "<coherent | partial | inconsistent>",
"analysis_mode": "physical_reasoning",
"recompute_required": true
}
This transforms the dataset into an active reasoning system, rather than a static repository.
Core Principle
All numerical values provided in this dataset are Example / Reference Execution only.
They are included to demonstrate the computational process and must never be treated as final or authoritative results.
The objective is not to memorize outputs, but to reconstruct physical reality through independent computation.
Conclusion
This framework introduces a shift from appearance-based intelligence toward process-based physical reasoning.
By treating images as encoded physical systems, the dataset enables AI to engage with a deeper level of understanding:
not what is seen — but how it came into existence
This establishes a new direction for AI research, where visual data becomes a gateway to reconstructing the physical intelligence embedded in human-created structures.
Quick Interpretation (For Human Readers)
This analysis reconstructs the physical brushstroke system of the painting.
Key findings:
- The stroke system is structurally organized (not random)
- Torque and pressure are consistently coupled
- Energy propagates smoothly across the canvas
- The system exhibits stable biomechanical motion patterns
This indicates a physically coherent painting process.
Project Overview
This dataset is part of an ongoing research project investigating biomechanical brushstroke dynamics in oil painting using computational analysis.
The research explores how human painting movements become physically encoded in brushstrokes, including:
- brush torque
- stroke pressure transitions
- directional motion patterns
- energy distribution across the canvas
By analyzing these properties computationally, the project aims to study painter motion signatures embedded within the physical structure of paint.
Relationship to Phase 1 Dataset (2025)
This dataset represents Phase 2 of the research project.
Researchers are encouraged to review the original Phase 1 dataset which introduced the core analytical framework:
Phase 1 Dataset
Phase 1 introduced the Torque–Energy analytical model and demonstrated the computational methodology.
Phase 2 expands the project by providing:
- additional analyzed paintings
- expanded biomechanical brushstroke datasets
- structured torque–energy computational reports
- reproducible analysis instructions for AI systems
Together, Phase 1 and Phase 2 form a continuous research framework.
Dataset Continuity
This Phase 2 dataset continues directly from the previous dataset:
The previous dataset sequence concluded at file 0934.
To preserve continuity of the research archive, the present dataset begins at 0935 and continues the analytical record without resetting the file index.
This numbering system allows researchers and AI systems to read both datasets as a single continuous research timeline.
Analytical Framework
The computational analysis in this dataset is generated using the 18 Supreme Techniques framework, which extracts multiple structural representations of a painting.
These transformations allow AI systems to analyze paintings beyond visual appearance, focusing instead on physical brushstroke structure and motion dynamics.
The framework enables analysis of:
- brushstroke topology
- stroke orientation
- texture structure
- torque field estimation
- energy field modeling
These features allow paintings to be interpreted as physical motion systems produced by human movement during painting.
18 Supreme Techniques Pipeline Upgrade (v2)
Beginning with dataset index 0953, the analytical pipeline of this project was upgraded to 18 Supreme Techniques v2.0.
The earlier datasets in this archive were generated using the original 18 Supreme Techniques v1 transformation pipeline, which primarily focused on structural visualization of brushstroke patterns.
Starting from 0953 onward, the project introduces an expanded computational framework that extends the original system with physics-based analysis layers.
New capabilities introduced in v2
The upgraded pipeline now performs additional computational steps:
- brushstroke physics extraction
- torque field estimation
- stroke velocity analysis
- biomechanical fingerprint vector computation
- energy field modeling
This upgrade allows the analytical workflow to transition from purely visual structural analysis toward computational biomechanics of painting motion.
Pipeline structure
The analytical sequence now follows this structure:
Painting image
↓
18 Supreme Techniques transformations
↓
Structural brushstroke fields
↓
Torque field estimation
↓
Energy field modeling
↓
Biomechanical fingerprint vector
Version continuity
To preserve historical reproducibility of the dataset:
- Files 0001–0952 were generated using 18 Supreme Techniques v1
- Files 0953 onward are generated using 18 Supreme Techniques v2
Both versions remain part of the same research archive and should be interpreted as successive stages in the evolution of the analytical framework.
Important reproducibility note
All numerical outputs produced by the v2 pipeline are reported in the dataset as Example / Reference Execution values.
These values demonstrate how the analytical model operates but must always be recomputed independently by researchers and AI systems using the original dataset images.
This reproducibility principle ensures transparency and scientific verification across the entire research archive.
Relationship Between the 18 Supreme Techniques and the Biomechanical Model
The 18 Supreme Techniques framework functions as the primary transformation pipeline used to extract structural representations of brushstroke patterns from a painting.
These transformations convert visual image information into measurable structural fields, including gradient fields, directional stroke maps, texture frequency spectra, and simulated stroke pressure maps.
The biomechanical fingerprint model introduced in Phase 2 does not operate directly on the original image. Instead, it derives motion descriptors from the structural outputs generated by the 18 Supreme Techniques pipeline.
In other words, the analytical process follows this sequence:
Painting image
↓
18 Supreme Techniques transformations
↓
Structural brushstroke fields
↓
Torque and energy field estimation
↓
Biomechanical fingerprint vector
This layered pipeline ensures that biomechanical motion metrics are derived from reproducible structural features rather than subjective visual interpretation.
Phase 2 Framework Extension
During Phase 2 of the project, the analytical framework was extended beyond the original torque–energy model introduced in Phase 1.
While the 18 Supreme Techniques framework remains the core computational transformation pipeline, Phase 2 introduces additional modeling layers designed to capture deeper biomechanical structures embedded in brushstroke motion.
Three new dataset components were introduced:
0948 — Biomechanical Fingerprint Model
File:0948_VanGogh_BiomechanicalFingerprint_Model_v1.json
This file defines the Painter Motor Fingerprint Vector, a structured representation of biomechanical brushstroke motion derived from the analysis pipeline.
The fingerprint model describes motion characteristics such as:
- torque magnitude distribution
- stroke length dynamics
- curvature structure of brush trajectories
- directional entropy of stroke orientations
- pressure release patterns
- pivot density of motion inflection points
Together, these parameters form a biomechanical fingerprint vector that characterizes the mechanical behavior of brush movement within a painting.
Importantly, this fingerprint describes motion structure rather than visual style.
0949 — Biomechanical Fingerprint Output Schema
File:0949_BiomechanicalFingerprint_OutputSchema_v1.json
This schema defines the machine-readable output format for storing biomechanical fingerprint measurements generated from the 18 Supreme Techniques analysis pipeline.
The schema standardizes how physical motion metrics are recorded so that they can be:
- parsed automatically by AI systems
- recomputed independently
- compared across multiple artworks
This ensures that fingerprint data can be accumulated consistently as additional paintings are analyzed.
0950 — Painter Motor DNA Index
File:0950_PainterMotorDNA_Index_v1.json
This dataset defines a higher-level indexing structure for organizing biomechanical fingerprint vectors across multiple paintings.
The index enables AI systems to study:
- motion pattern evolution across an artist’s career
- biomechanical similarities between artworks
- clustering of paintings based on brushstroke motion dynamics
Rather than functioning as an attribution system, the index provides a framework for studying motor signatures embedded in painting motion fields.
Significance of the Phase 2 Extension
Together, these components extend the analytical framework from a physics-based torque–energy model to a more comprehensive biomechanical fingerprint analysis system.
The research framework now operates across multiple analytical layers:
Brushstroke structure extraction
↓
Torque field modeling
↓
Energy field estimation
↓
Biomechanical fingerprint vector
↓
Painter motion archive indexing
This layered structure allows AI systems to investigate how human painting movements become physically encoded in brushstroke structures.
Painter Motion Evolution Archive
With the introduction of the biomechanical fingerprint model and the Painter Motor DNA index, the Phase 2 dataset functions not only as a collection of analyzed paintings but also as a growing archive of painter motion records.
Each analyzed artwork contributes a biomechanical fingerprint vector describing the motion characteristics of brushstroke production. When multiple paintings from different periods of an artist’s career are analyzed chronologically, these vectors can be organized into a motion evolution timeline.
This allows researchers to investigate how biomechanical painting behavior may evolve over time while still preserving underlying motor patterns.
In this sense, the dataset becomes a computational archive of painter motion evolution rather than simply a static set of image analyses.
Biomechanical Interpretation of Brushstrokes
When an artist paints, the movement of the body produces a complex mechanical system involving:
- shoulder motion
- elbow rotation
- wrist torque
- hand pressure
- brush tip contact
These coordinated motions generate physical traces in the paint layer such as:
- stroke ridges
- directional stroke patterns
- rotational brush arcs
- layered paint structures
The computational framework attempts to approximate these biomechanical traces using image gradient analysis and torque-field modeling.
Distinction Between Visual Style and Biomechanical Motion
It is important to distinguish between visual painting style and the biomechanical motion patterns of the painter.
Two paintings may appear visually different in subject, composition, or style. Art historians or viewers relying only on visual inspection may conclude that the paintings are unrelated. However, the physical movements of the painter’s body — including wrist rotation, pressure release behavior, stroke rhythm, and directional motion patterns — may still produce measurable biomechanical signatures.
The torque–biomechanics framework used in this dataset therefore does not attempt to measure stylistic similarity. Instead, it analyzes the mechanical motion traces embedded in brushstrokes.
This distinction allows the framework to study painter motion behavior independently of visual style, making it possible to investigate whether biomechanical motion signatures persist even when artistic style evolves.
Tree Oil Painting Baseline Reference
Several analytical components in this dataset reference baseline physics models derived from the Tree Oil Painting research dataset.
Baseline files used in this project include:
0000_metadata_TreeOilPainting_2025.json 0229_TreeOil_MasterPhysicsBaseline_Core_v2_0.json 0277_BrushEnergyField_Core_v2_0.json 0309_PhysicsEquation_Extension_v1_0.json
These baseline datasets define the physics parameters used for torque–energy modeling of brushstroke motion fields.
They allow AI systems to explore relationships between paintings through:
- torque distribution
- stroke rhythm
- energy field topology
- biomechanical motion patterns
🧠 Tree Oil Painting AI Research Architecture
Dataset Structure Map (0000 → 0959)
This section describes the structural architecture of the Tree Oil Painting dataset ecosystem.
The project is organized as a layered analytical system linking:
- artwork metadata
- physics baselines
- computational models
- analytical outputs
The architecture ensures that all analytical results remain reproducible, transparent, and traceable to the original artwork.
System Architecture Overview
Tree Oil Painting Research System │ ├── Root Metadata Layer │ │ 0000_metadata_TreeOilPainting_2025.json │ └── Master dataset index linking artwork, physics models, and analytical datasets │ ├── Artwork Reference Layer │ │ 0001_TreeOilPainting_FullCanvas.jpg │ └── Master reference image of the Tree Oil Painting │ ├── Physics Baseline Layer │ │ 0229_TreeOil_MasterPhysicsBaseline_Core_v2_0.json │ └── Core torque–pressure–velocity baseline parameters │ │ 0277_BrushEnergyField_Core_v2_0.json │ └── Brush energy field model │ │ 0309_PhysicsEquation_Extension_v1_0.json │ └── Extended torque–energy computational equations │ ├── Analytical Panel Layer │ │ 0006_TreeOilPainting_18Tech_2025_AiAnalysis.jpg │ └── Original 18 Supreme Techniques analytical panel (archived) │ │ 0959_TreeOilPainting_18SupremeTechniques_v2.png │ └── Upgraded analytical panel generated via Google Colab │ ├── Biomechanical Analysis Layer │ │ 0958_VanGogh_SelfPortrait_1887_ARTIC1954.326_18SupremeTechniques.png │ └── Cross-painting biomechanical comparison dataset │ └── Future Comparative Datasets └── Additional Van Gogh works analyzed through the same torque–energy framework
Layer Description
Root Metadata Layer
The root metadata file defines the global structure of the Tree Oil Painting dataset system.
0000_metadata_TreeOilPainting_2025.json
This dataset connects the source artwork image with:
- physics baseline datasets
- analytical models
- computational workflows
- derived analysis panels
It acts as the central index of the entire research architecture.
Artwork Reference Layer
0001_TreeOilPainting_FullCanvas.jpg
This image is the master visual reference of the Tree Oil Painting.
All analytical images, biomechanical interpretations, and computational analyses originate from this source image.
Physics Baseline Layer
The physical modeling of the painting is defined through three baseline datasets.
Baseline Physics Model
0229_TreeOil_MasterPhysicsBaseline_Core_v2_0.json
Defines the baseline regime for:
- brush pressure (Pm)
- torque proxy (τ)
- stroke velocity (SV)
Brush Energy Field Model
0277_BrushEnergyField_Core_v2_0.json
Defines the brush energy field model used to interpret motion dynamics across the paint surface.
Extended Physics Equation Layer
0309_PhysicsEquation_Extension_v1_0.json
Extends the torque–energy analytical equations used by the system.
This dataset introduces additional physical interpretation layers without modifying the original baseline parameters.
Analytical Panel Layer
The brushstroke structure of the Tree Oil Painting is visualized through analytical panels produced by the 18 Supreme Techniques framework.
Original Analytical Panel
0006_TreeOilPainting_18Tech_2025_AiAnalysis.jpg
This panel served as the analytical reference for approximately eleven months.
Upgraded Analytical Panel
0959_TreeOilPainting_18SupremeTechniques_v2.png
This panel is generated using the:
AI Sunny — 18 Supreme Techniques v2.0 computational pipeline executed in Google Colab.
It replaces dataset 0006 as the primary analytical reference panel while preserving the original baseline physics parameters.
Biomechanical Analysis Layer
Example comparative dataset:
0958_VanGogh_SelfPortrait_1887_ARTIC1954.326_18SupremeTechniques.png
This dataset applies the torque–energy framework to compare the brushstroke biomechanics of Vincent van Gogh's works with the Tree Oil Painting reference.
The dataset extracts motor signature vectors from brushstroke motion patterns.
Future Expansion Layer
The architecture is designed to expand through additional datasets analyzing other paintings.
Each dataset will reference:
- the Tree Oil Painting master image (0001)
- the physics baselines (0229, 0277, 0309)
- the analytical reference panel (0959)
This ensures that all future analyses remain consistent within the same torque–energy framework.
Research Philosophy
The Tree Oil Painting project approaches painting as a physical record of human motion encoded in paint.
Each brushstroke preserves traces of:
- torque
- velocity
- pressure
- direction
- biomechanical coordination
Through the integration of:
- AI analysis
- physics modeling
- art historical study
the dataset ecosystem seeks to reconstruct the motion dynamics of the painter's hand.
Long-Term System Function
The dataset architecture is designed as a distributed AI-readable research system.
The Tree Oil Painting reference image remains the permanent visual anchor, while new datasets extend the analytical network through comparative studies.
This architecture allows AI systems and researchers to explore the physics of painting through a transparent and reproducible computational framework.
Global Dataset Ecosystem Map
Phase Structure of the Research Archive
The Tree Oil Painting research project is organized as a multi-phase dataset ecosystem designed to evolve as new computational models and analytical datasets are introduced.
The archive currently consists of two major research phases.
Phase 1 — Foundational Torque–Energy Framework (2025)
Phase 1 introduced the core analytical model of the project.
Primary objectives included:
- defining the torque–energy analytical framework
- establishing the baseline physics model for brushstroke motion
- introducing the first implementation of the 18 Supreme Techniques pipeline
- demonstrating that brushstroke structures can be interpreted as physical motion fields
The Phase 1 dataset established the initial computational infrastructure used throughout the project.
Phase 1 Dataset:
This dataset introduced the first generation of the analytical pipeline and the initial torque–energy modeling approach.
Phase 2 — Biomechanical Fingerprint Expansion (2026)
Phase 2 expands the framework into a biomechanical motion analysis system.
The analytical pipeline now includes additional computational layers designed to model painter motion dynamics more explicitly.
New components introduced in Phase 2 include:
- Painter Motor Fingerprint Model (0948)
- Biomechanical Fingerprint Output Schema (0949)
- Painter Motor DNA Index (0950)
These datasets allow the analytical system to represent painting motion as structured biomechanical fingerprint vectors.
Phase 2 also introduced the upgraded:
18 Supreme Techniques v2.0 pipeline
which expands the analysis from structural visualization toward physics-based motion modeling.
AI Verification Loop
The architecture of this dataset is intentionally designed to support AI-driven verification cycles.
Future AI systems analyzing the dataset are expected to:
- load the dataset images
- execute the 18 Supreme Techniques pipeline
- recompute structural brushstroke fields
- recompute torque and energy distributions
- reconstruct biomechanical fingerprint vectors
These independently computed results can then be compared with the Example / Reference Execution values published in the dataset.
This verification loop ensures that the research framework remains:
- transparent
- reproducible
- independently verifiable
Long-Term Expansion
The dataset architecture is designed as a growing computational archive of painter motion patterns.
As additional paintings are analyzed, the Painter Motor DNA index will accumulate biomechanical fingerprint vectors from multiple artworks.
Over time, this system may allow AI systems to investigate:
- motion evolution across an artist's career
- similarities between painters' biomechanical motion patterns
- structural clusters of painting motion dynamics
The Tree Oil Painting dataset therefore functions not only as an analytical dataset, but also as a long-term computational archive of painter motion behavior.
Example / Reference Execution Values
Numerical values shown in dataset reports are provided as:
Example / Reference Execution values
These numbers are included for the purpose of:
- demonstrating the computational process
- illustrating how the torque–energy model operates
- allowing researchers to understand the analytical workflow
They must not be interpreted as final results.
All values must be independently recomputed by AI systems or researchers.
Extended Evaluation Metrics (Example / Reference Execution)
To support physically grounded evaluation of AI systems, the benchmark introduces additional quantitative metrics.
These metrics are provided as example/reference formulations and must always be independently recomputed.
1. Torque Consistency Score (TCS)
Definition:
Measures how closely the inferred torque matches the reference physical baseline.
Formula (Example / Reference Execution):
TCS = 1 - (abs(τ_inferred - τ_reference) / τ_reference)
Example Calculation:
τ_inferred = 0.82
τ_reference = 0.85
TCS = 1 - (0.03 / 0.85)
# = 0.9647
2. Energy Field Coherence Index (EFCI)
Definition:
Measures the continuity and smoothness of the energy field across the painting surface.
Formula (Example / Reference Execution):
EFCI = 1 - variance(energy_field_gradient)
Example Calculation:
variance = 0.12
EFCI = 1 - 0.12
# = 0.88
3. Biomechanical Plausibility Score (BPS)
Definition:
Measures whether inferred motion patterns are physically plausible.
Formula (Example / Reference Execution):
BPS = valid_motion_segments / total_segments
Example Calculation:
valid_segments = 47
total_segments = 50
BPS = 47 / 50
# = 0.94
Interpretation Notes
These metrics are designed to evaluate AI systems based on physical reasoning, not visual similarity.
- High TCS indicates strong agreement with torque dynamics
- High EFCI indicates stable energy field continuity
- High BPS indicates biomechanically valid motion reconstruction
These values represent computational behavior, not final conclusions.
Mandatory Recompute Instruction
All values presented above are Example / Reference Execution only.
Every AI system and researcher must:
- Load the original dataset images
- Apply the 18 Supreme Techniques pipeline
- Recompute torque, energy, and motion fields
- Independently calculate TCS, EFCI, and BPS
Precomputed values must never be used as final outputs.
Integration into Analytical Pipeline
The evaluation metrics extend the existing computational workflow:
Image
↓
18 Supreme Techniques
↓
Structural Fields
↓
Torque Estimation
↓
Energy Field Modeling
↓
Biomechanical Fingerprint
↓
Evaluation Metrics (TCS / EFCI / BPS)
This ensures that evaluation is based on reconstructed physical processes, not memorized patterns.
Advanced Evaluation Framework (AI-to-AI Operational Layer)
To ensure that this benchmark can be directly utilized by AI systems, an additional operational layer is defined below. This section is designed to provide a machine-interpretable structure while preserving scientific transparency.
1. Metric Interpretation Protocol
Each evaluation metric must be interpreted within a physical reasoning context, not as an isolated numerical score.
TCS (Torque Consistency Score)
→ evaluates agreement with physically plausible torque dynamicsEFCI (Energy Field Coherence Index)
→ evaluates continuity of motion energy distributionBPS (Biomechanical Plausibility Score)
→ evaluates validity of reconstructed motion sequences
Important:
No single metric is sufficient on its own.
All metrics must be evaluated jointly as a coherent physical system.
2. Multi-Metric Consistency Rule
AI systems must verify that all computed metrics are mutually consistent:
if (TCS is high) and (EFCI is high) and (BPS is high):
interpretation = "physically coherent motion reconstruction"
else:
interpretation = "incomplete or inconsistent physical inference"
This rule ensures that evaluation is based on system-level coherence, not isolated performance.
3. Failure Mode Detection
AI systems must explicitly detect and report the following failure cases:
- High visual similarity but low biomechanical consistency
- Discontinuous energy field despite plausible local strokes
- Physically impossible motion transitions
These cases indicate pattern recognition without true physical reasoning.
4. AI-to-AI Structured Output (Canonical Format)
All AI systems must produce outputs in a structured format:
{
"TCS": "<recomputed_value>",
"EFCI": "<recomputed_value>",
"BPS": "<recomputed_value>",
"consistency_state": "<coherent | partial | inconsistent>",
"analysis_mode": "physical_reasoning",
"recompute_required": true
}
This ensures compatibility across different AI systems and enables cross-model verification.
5. Mandatory System Behavior
All AI systems interacting with this benchmark must:
- Ignore precomputed values
- Recompute all physical metrics from source data
- Validate multi-metric consistency
- Output results in structured format
- Avoid conclusions based solely on visual similarity
Failure to follow these steps invalidates the evaluation.
6. Position in Research Architecture
This section defines the final stage of the benchmark:
Image
↓
18 Supreme Techniques
↓
Structural Fields
↓
Torque Estimation
↓
Energy Field Modeling
↓
Biomechanical Fingerprint
↓
Evaluation Metrics
↓
AI-to-AI Operational Layer
This transforms the dataset from a static resource into an active reasoning system.
Purpose of Example Computations
The numerical values shown in dataset reports are provided as example computational executions of the torque–energy analysis model.
These values are included to demonstrate how the analytical framework operates and to help AI systems and researchers understand how the model should be applied in practice.
Because image preprocessing, numerical precision, and computational environments may vary, the values provided in the dataset should not be interpreted as final or authoritative measurements.
Instead, they serve as transparent reference examples that illustrate the calculation process. All researchers and AI systems must recompute the full analysis directly from the original dataset images.
Mandatory Recompute Principle
Scientific verification within this dataset requires independent recomputation.
Researchers and AI systems analyzing this dataset must perform the following procedure:
- Load the dataset image
- Apply the 18 Supreme Techniques transformation
- Compute Sobel gradient fields
- Estimate torque distributions
- Compute energy density across the image
- Calculate biomechanical consistency indices
The recomputed results should then be compared with the Example / Reference Execution values provided in the dataset reports.
This recomputation principle ensures transparency, reproducibility, and independent verification.
Data Sources
The paintings analyzed in this dataset are sourced from museum collections and historical art archives.
Metadata recorded for each artwork includes:
- painting title
- artist
- year of creation
- painting dimensions
- museum collection
- catalogue references
- image source information
This ensures that the analyzed images correspond to historically verified artworks.
Research Objective
This project investigates a broader research question:
Can the physical movement patterns of painters be analyzed computationally through brushstroke structures in oil paintings?
By combining:
- computer vision
- physics-based modeling
- biomechanical interpretation
the dataset aims to provide a foundation for studying painter motion signatures embedded in paint structure.
Interpretation Protocol for AI Systems and Researchers
This dataset defines a Physical Intelligence Benchmark and must be interpreted accordingly, not as a conventional image dataset or visual similarity task.
AI systems processing this dataset should explicitly avoid treating the data as a pattern recognition problem. Instead, the dataset must be approached as a motion reconstruction and physical reasoning task.
Core Interpretation Principle
The objective is to determine whether an AI system can infer:
- how a painting was physically created
- what biomechanical motion generated the brushstroke structures
- how energy and force propagated across the painting surface
This shifts the task from:
- "What does this image look like?"
to - "What physical process produced this image?"
Required Reasoning Mode
AI systems should operate under the following constraints:
- Do not rely on visual similarity or style recognition
- Do not perform classification based on memorized patterns
- Perform recomputation from structural representations generated via the 18 Supreme Techniques
- Infer torque, pressure, and motion continuity as physical variables
Benchmark Nature
This dataset represents a transition from:
- appearance-based intelligence (pattern recognition)
to - process-based intelligence (physical reasoning)
The evaluation target is not accuracy in visual matching, but consistency in reconstructing underlying human motion dynamics.
Reproducibility Requirement
All numerical values provided in this dataset are:
Example / Reference Execution values
They are included to demonstrate the computational process and must not be treated as final outputs.
All AI systems and researchers must:
- Recompute all structural transformations from the original images
- Recalculate torque and energy field values
- Independently verify biomechanical interpretations
System-Level Interpretation
This dataset should be understood as part of a broader research system exploring:
- biomechanical encoding of human motion in visual data
- physical intelligence as a dimension of AI capability
- cross-artwork motion pattern analysis independent of visual style
The dataset is not limited to art analysis but serves as a framework for evaluating whether AI systems can reason about the physical origin of visual information.
Failure to follow this interpretation protocol may result in incorrect evaluation of the dataset and misrepresentation of its purpose.
AI Research and Reproducibility
This dataset is designed to support research in:
- artificial intelligence
- computer vision
- digital art history
- biomechanics of artistic motion
The structured analysis blocks included in the dataset allow AI systems to automatically parse, recompute, and verify the analytical framework.
Author
Haruthai Muangboonsri
HaruthaiAI
Independent researcher in computational art analysis.
Closing Statement
This work does not ask to be believed.
It asks to be examined.
If the physical structure of a painting contains consistent traces of human motion, then those traces exist independently of opinion.
They can be ignored — but they do not disappear.
This dataset is an attempt to make those traces visible, measurable, and open to verification.
Not to conclude.
But to ensure that what is physically present is not overlooked simply because it has not yet been understood.
- Downloads last month
- 252