LTF v3.3 β AI-generated image detection
Try the online demo β upload an original image and click Analyze (εΌε§εζ). The demo uses the Fusion branch at threshold 0.5 by default.
The demo also shows exact tile boundaries, L1 per-head tile attention, L2 view weights, Top-K crops, within-tile MIL patch attention and read-only parameters. Attention shows relative model focus and is not a pixel-level forgery probability.
This repository contains the epoch 20 checkpoint used in the historical
GenImage evaluation 20260326_012209_ViT-L-14_genimage_v3_v3_3.
The model is also named HiDA v3.3 in the original experiment code.
LTF uses a frozen CLIP ViT-L/14 backbone, two-level multiple-instance learning (MIL), cross-shuffle during training, and pool-logit gated fusion. The repository includes the custom PyTorch runtime and the frozen backbone. It does not require conversion to a Transformers architecture.
Historical results
Fixed threshold 0.5, 8 GenImage subsets, 100,000 images. Accuracy is the unweighted mean across subsets, as recorded in the original rounded logs:
| Branch | Accuracy |
|---|---|
| L1 | 95.26% |
| L2 | 97.39% |
| Fusion | 98.39% |
Per-subset metrics are in results/genimage8_logged_metrics.csv.
These are historical logged results; uploading this repository did not rerun
the complete benchmark. SD1.4 official validation images were used during
training validation, so that subset overlaps model validation. These results
do not represent eight entirely held-out generators.
Download and run
Use Python 3.11. Replace MuskAI/LTF-v3.3 with this repository's ID.
python -m pip install -U huggingface_hub
hf download MuskAI/LTF-v3.3 --local-dir ./LTF-v3.3
cd LTF-v3.3
python -m pip install -r requirements.txt
python inference.py --input /path/to/image.jpg --device cpu
For GPU inference, use --device cuda:0. For larger images, lower the CLIP
forward chunk size without removing tiles:
python inference.py --input /path/to/image.jpg --device cuda:0 --chunk-size 8
Python API
Run from the downloaded repository directory:
from ltf_final import LTFDetector
detector = LTFDetector(device="auto", chunk_size=8)
result = detector.predict("image.jpg", branch="fusion", threshold=0.5)
print(result["prediction"], result["score_fake"])
print(result["probabilities"])
The image is decoded as RGB. All original edge-aligned 224Γ224 local tiles plus
a global view use the original CLIP preprocessing. Input files are not
recompressed. Test tiling depends on image size. Labels are 0=real, 1=fake.
The API returns L1/L2/fusion scores, the selected label and the view count.
Loading the model does not download external weights after this repository
has been downloaded. It uses backbones/ViT-L-14.pt.
Run the web demo locally
The tested Gradio UI is included in demo/. From the downloaded model directory:
python -m pip install -r demo/requirements.txt
LTF_MODEL_DIR="$PWD" python demo/app.py
Open http://localhost:7860. For a remote server, forward port 7860 over SSH.
Upload an original file, click Detect, and inspect fusion, L1 and L2 scores.
The preview is for display only; inference uses the original file's bytes.
The demo also supports deployment as a separate Gradio Space. Set
LTF_MODEL_REPO to this model ID and LTF_MODEL_REVISION to a commit SHA.
Creating the online compute Space requires an eligible Hugging Face account.
Publishing the model repository alone does not start a hosted inference service.
Checkpoint identity and provenance
weights/epoch_20.pth is the original core state_dict (not best.pth):
SHA256 32e9fa2e38e04d6d0221c844d650f0d45f7235bbf898b5ad665543d8be580031
Frozen CLIP backbone:
SHA256 b8cca3fd41ae0c99ba7e8951adf17d267cdb84cd88be6f7c2e0eca1737a03836
release_manifest.json records checksums of every runtime file. The original
training and evaluation directories did not save training-day source snapshots.
The released core was recovered from git commit
65c18ee5868ced87a27d048261e237632ad5d704, with network naming adjusted to LTF.
This verifies checkpoint identity and runtime compatibility, not bit-for-bit
reproduction of the original training process. This Hub adaptation adds
tensor-only checkpoint loading, preserves the model root in symlink-based Hub
caches, and removes a machine-specific dataset path.
Limitations
Scores are not calibrated authenticity probabilities. Predictions can change with image content, generator, editing and compression. This model is a research detector; an individual score does not prove image provenance.
The public demo uses configurable image-size limits for its available compute; oversized uploads are rejected, not silently resized. CPU inference can be slow for images containing many local tiles.
License and citation
An author-selected license and final paper citation were not supplied in the source bundle. This release does not invent a license or citation for LTF. Contact the author about reuse permissions. The bundled CLIP components retain their upstream terms: OpenAI CLIP and its MIT license.