multi-provider-source-v7)This repository contains the unified (55.53 MB), CPU-only, pickle-free (allow_pickle=False) multimodal weights for the abstaining remove-ai-watermarks pixel source-pipeline and synthetic-image classifier across 10 AI provider families (openai, google, bytedance, flux, stable_diffusion, meta, midjourney, ideogram, firefly, xai) plus unknown, with Stage-2 model/export version attribution (version_subfamily) and calibrated binary synthetic detection (synthetic_detected, synthetic_probability) under foreign resize and container transcoding.
This model is not a SynthID detector and does not decode or verify any cryptographic or invisible watermark payload (watermark_truth="unknown"). Its labels describe complete generation and export pixel pipelines, so renderer, geometry, encoding, post-processing, and watermark state may be confounded. unknown does not mean that SynthID is absent, and a provider label does not prove that SynthID is present.
pip install "remove-ai-watermarks[source-classify]"
import remove_ai_watermarks as raiw
res = raiw.classify_source("image.jpg")
print(res.label, res.version_subfamily, res.synthetic_detected, round(res.synthetic_probability, 4))
# e.g. 'openai' 'openai:gpt_image_1_5' True 0.9942 (native AI export)
# e.g. 'unknown' None True 0.9412 (foreign 1024x1024 transcoded AI image)
# e.g. 'unknown' None False 0.0084 (camera photo, UI screenshot, diagram, or scanned document)
Command line (installed CLI or zero-setup uv run with the bundled infer.py and open-detector benchmark.py):
remove-ai-watermarks classify-source image.jpg --json
uv run https://huggingface.co/wiltodelta/raiw-source-classify/resolve/main/infer.py image.jpg --json
uv run https://huggingface.co/wiltodelta/raiw-source-classify/resolve/main/infer.py --inspect-weights
uv run https://huggingface.co/wiltodelta/raiw-source-classify/resolve/main/benchmark.py
uv run https://huggingface.co/wiltodelta/raiw-source-classify/resolve/main/benchmark.py --tier frontier
Try it in your browser on the Remove AI Watermarks Hugging Face Space or at raiw.cc. Set RAIW_SOURCE_CLASSIFY_WEIGHTS to multi-provider-source-v7.npz or its parent directory for air-gapped deployments.
512-d, extract_512d_v4_features):112-d adaptive overlapping native-tile truncated SRM micro-residuals across flat and textured regions.128-d phase-resolved modulo-8 / modulo-16 sublattice energy, cropped-grid secondary-phase detection, interior VAE modulation, and 8x8 block-DCT quantization comb statistics.144-d Moisan periodic-plus-smooth (p + s) spectral decomposition, radial decay profile, and continuous scale-invariant period-comb scanner.96-d patch-averaged cross-channel spectral phase coherence, Bayer/Quad-Bayer CFA periodicity, heteroscedastic shot-noise slope, and radial chromatic-aberration alignment.32-d explicit codec, aspect-ratio, scale, and social-recompression regime vector (x[480:512]).ViT-S/16 @ 384 weight-fp16 ONNX, vit_onnx_bytes):ViT-S/16 @ 384 frozen visual stem (OwensLab/commfor-model-384 / TensorFold/detectra-v1 lineage) paired with open-world multi-task readout heads ([cls_tok(384), patch_mean(384), patch_std(384)] -> 28-d vector [orig_syn(0), prov_10(1:11), foreign_ai(11), real_non_ai(12), gfx_veto(13), dom_syn(14), emb(15:28)]) stored as a weight-fp16 -> fp32-compute ONNX graph inside the pickle-free .npz bundle and executed via onnxruntime on CPU, plus a conditional short_edge=672 2-crop gray-zone pass for widescreen native exports.154 unit channels + 96 2D gates + 82-d open-set provider head + 67-d open-world synthetic fusion):512-d ensemble in pure NumPy (5 multi-seed Regime-FiLM LayerNorm + SiLU residual MLP trunks + 65 packed HistGradientBoosting specialist forests) feeding an 11-channel regime-standardized acceptance detector, conditional attribution MLP, and an 82-d -> 64-d -> 12 open-set multimodal provider head (10 providers vs logsumexp(foreign_ai, real_non_ai)) that recovers exact label on transcoded 10-provider images while keeping unseen generators as honest unknown.67-d -> 48-d -> 1 LayerNorm + SiLU skip multimodal synthetic fusion head combining the 28-d ViT output, bounded 11-channel Stage-1.5 physics acceptance vector, positive/negative global physics cues, the 26-d codec-invariant born-digital graphics score, and the 32-d regime block. When container laundering or an unseen generator prevents 10-family attribution, label honestly abstains as unknown while synthetic_detected and synthetic_probability retain binary synthetic recall without sacrificing screenshot, diagram, or document specificity.9 stacked blended version heads, 666-d):HistGradientBoosting tree ensembles blended with standardized linear heads over the 512-d descriptor concatenated with the 154 Stage-1 unit scores (666-d) to predict the canonical model/export subfamily (version_subfamily).multi-provider-source-v7.npz (allow_pickle=False, 55.53 MB)4faef51eea48f8194ee804fab1782c48d626384e115e2ef70884f8e8393d7eb1The artifact contains only learned parameters, weight-fp16 ONNX graph bytes, normalization constants, tree arrays, labels, and calibrated thresholds. It does not contain training images, paths, image hashes, embeddings, or a training catalog.
benchmark.py)Evaluated end-to-end via classify_source on 200 non-AI false-positive traps (50 camera photos, 50 UI screenshots, 50 science diagrams, 50 scanned documents), 125 untouched native AI exports (25 each across FLUX.1-dev, GPT-Image-1, Gemini Nano Banana Pro, SDXL, Midjourney v6.1), and 253 square-resized/transcoded crowd-arena AI images (Rapidata and bitmind 1024x1024 JPEG/WebP/PNG):
200 traps): 0.0% (0 / 200) across all four slices (0% camera photo, 0% UI screenshot, 0% science diagram, 0% scanned document) and 100.0% precision.125): 99.2% binary synthetic recall (124 / 125, including 96.0% on widescreen 1408x768 Gemini Nano Banana Pro and 100.0% on widescreen 1456x816 Midjourney v6.1, FLUX.1-dev, GPT-Image-1, and SDXL) and 71.2% exact generator attribution (89 / 125).253): 87.0% binary synthetic recall (220 / 253, up from 30.4% in v6) and 18.6% exact generator attribution (47 / 253, up from 9.5% in v6 via open-set multimodal attribution while keeping honest unknown abstention when provider identity is unrecoverable; 329 ms median single-image CPU latency).The operating point was selected strictly on parent-disjoint calibration partitions (parent_sha256 grouped). Across sealed held-out benchmarks:
4,415 resolved locked_test parents (2,074 supported AI + 2,341 real non-AI across 15 real varieties): 0.09% real synthetic FPR (2 / 2,341 false positives, 0 on camera photos, UI, diagrams, vectors, and documents; 0.04% real provider FPR, 1 / 2,341), 97.54% native synthetic recall (92.82% exact provider recall; 0.9995 ROC-AUC / 0.9994 PR-AUC), and 65.43% synthetic recall (25.41% exact provider recall) after forced 1024x1024 JPEG/WebP arena transcoding across all 10 provider families (0.9740 ROC-AUC).84 held-out views outside the 10 provider families): 97.62% honest label="unknown" abstention (82 / 84; 0 / 58 false attributions on cal-fit foreign AI) with 69.05% binary synthetic recall across unseen generators including hard-flash DreamBooth style transfers.selective-v8 end-to-end RGB + ViT (1,512 untouched real-variety parents): 0 / 1,297 (0.00% FPR) across all camera photo (0 / 25), diagram (0 / 305), document (0 / 308), interface (0 / 327), and vector (0 / 332) parents, and 4 / 215 on game_art (4 / 1,512 = 0.26% overall synthetic FPR; 0 / 1,512 = 0.00% provider FPR; 1 / 2,267 = 0.04% across all 2,267 512-d rows).10,585-row locked_test 512-d provider router: 99.93% unknown specificity (3 / 4,538 false positives, 0.07% FPR), 83.28% exact provider recall (92.45% native, 92.95% resized, 77.06% social-recompressed), 99.94% any-AI precision, and 95.93% Stage-2 version accuracy given exact provider.| Provider | Test count | Standalone ROC AUC | Router recall | Router precision | Stage-2 version accuracy |
|---|---|---|---|---|---|
openai |
1,478 | 0.9997 | 92.1% | 96.1% | 95.6% |
google |
1,382 | 0.9996 | 86.4% | 97.5% | 97.0% |
bytedance |
389 | 0.9863 | 72.0% | 98.9% | 96.8% |
flux |
1,357 | 0.9993 | 80.8% | 95.2% | 95.0% |
stable_diffusion |
955 | 0.9987 | 77.2% | 98.1% | 98.0% |
meta |
94 | 0.9995 | 86.2% | 98.8% | - (meta:imagine) |
midjourney |
123 | 1.0000 | 82.1% | 87.8% | 98.0% |
ideogram |
118 | 0.9999 | 84.7% | 78.7% | 77.0% |
firefly |
103 | 0.9948 | 61.2% | 91.3% | 100.0% |
xai |
48 | 0.9990 | 45.8% | 91.7% | 77.3% |
Install remove-ai-watermarks[source-classify] and call the explicit Python API classify_source(path) or CLI command remove-ai-watermarks classify-source <IMAGE>. identify does not run this model. Set RAIW_SOURCE_CLASSIFY_WEIGHTS to the .npz file or its parent directory for offline deployments.
multi-provider-source-v7.npz 58 MBmulti-provider-source-v6.npz 16 MB