language	en
library_name	negate
license_name	MPL-2.0 + Commons Clause 1.0
compatibility	macos windows linux

negate
_{critical analysis of origin detection}

A scanning, training, and research library for scrutinizing illustration origin detection methods.

About

Negate is a modular system of image processing and feature extraction pipelines that measure machine aptitude of differentiating between synthetic and human-origin illustrations.

Quick Start

^{Terminal Powershell sh}

1.

2. uv tool install 'negate @ git+https://github.com/darkshapes/negate'

3. negate infer image.webp

Result	Translated Source
SYN	Synthetic/AI
GNE	Genuine/Human
?	High Uncertainty

Tip

To run without installing, use the uv command uvx --from 'negate @ git+https://github.com/darkshapes/negate' infer image.webp

Training

Train a new model with the following command:

negate train

Tip

type a path to an image file or directory of image files to add genuine human origin assets to the dataset add synthetic images using -s before a path

Technical Details & Research Results

Abstract

Previous research has demonstrated the possibility of identifying deepfakes, synthetic images, illustrations and photographs. Yet generative models have since undergone dramatic improvements, challenging past identification research and calling into question the future efficacy of these developments. Most methods chose images easily discernible as synthetic by the naked eye of a trained artist, or evaluated their success against open models exclusively. In this work, we create a comprehensive analysis suite for decomposition and feature extraction of digital images to study the effectiveness of these methods. Then, using an ensemble of previous techniques, we train simple decision trees and SVM models on these features to achieve >70% accuracy in detecting synthetic vs. genuine illustrations. Our methods of training and inference require only consumer-grade hardware, use exclusively consensual datasets provided by artists and Creative-Commons sources, and provide reliable estimates against the modern image products of both open and black-box closed-source models.

Included Methods

• local binary pattern

• gray lvl co-occurrence matrix

• energy

• complexity

• microtexture

• histogram oriented gradient (hog)

• variance

• kurtosis

• skew

• palette features

• spectral features

• haar wavelet

• laplacian

• gaussian diff

• snr/noise entropy

• random resize crop

• patchification

• l1/mse/k1/bce

Feature Processing Options

• Decision Tree + PCA

• SVM (RBF)

• MLP

• LR

Structure

Directories are located within $HOME\.local\bin\uv\tools or .local/bin/uv/tools

Data	Location	source
adjustable parameters	config/config.toml	included
downloaded datasets	.datasets/	HuggingFace
downloaded models	/models root folder	HuggingFace
trained models	/models date-numbered subfolders	generated via training
training metadata	/results date-numbered subfolders	generated via training

---

Module	Summary	Purpose
negate	core module	Root source code folder. Creates CLI arguments and interprets commands.
→→ decompose	image processing	RRC, Wavelet Transform - arxiv:2511.14030 https://arxiv.org/abs/2504.07078
→→ extract	feature processing	Residual analysis, VIT/VAE extraction, cross‑entropy loss - arxiv:2411.19417
→→ io	load / save / state	Hyperparameters, image datasets, console messages, model serialization and conversion.
→→ metrics	evaluation	Graphs, visualizations, model performance metadata, and a variety of heuristics for results interpretation.
→ inference	predictions	Detector functions to determine origin from trained model predictions.
→ train	XGBoost	PCA data transforms and gradient-boosted decision tree model training.

@misc{darkshapes2026,
    author={darkshapes},
    title={negate},
    year={2026},
    primaryClass={cs.CV},
    howpublished={\url={https://github.com/darkshapes/negate}},
}