Real-or-fake image check
According to the model card, the model checks whether an image is real or fake, meaning AI-generated.
Open Source Model Profile · dima806
ai_vs_real_image_detection
ai_vs_real_image_detection is an 85.8M-parameter ViT image classifier from dima806. According to the model card, it checks whether an image is real or AI-generated, with an explicit warning about concept drift.
- Publisher
- dima806
- Task
- image-classification
- Model type
- vit
- License
- apache-2.0
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
ai_vs_real_image_detection is published by dima806 as an image-classification model. The captured configuration identifies ViTForImageClassification with a vit model type, and Safetensors metadata reports 85800194 parameters. Hub tags associate it with google/vit-base-patch16-224-in21k as base and fine-tune, while the card describes real-versus-fake image checking.
Recorded capabilities
ViT base lineage tags
Hub tags list google/vit-base-patch16-224-in21k as both base model and fine-tune, with a captured ViTForImageClassification config.
Explicit drift warning
The publisher warns the training data is about two years old, notes significant concept drift from newer deepfake generators, and recommends retraining over threshold adjustment.
Kaggle training reference
The card points to a Kaggle notebook for further CIFAKE ViT training details.
Use cases in the source record
- Real-versus-AI-generated image screening experiments using the documented binary classification behavior.
- Retraining studies that use newer labeled data to address the publisher-reported drift in generative-image quality.
Limitations and unknowns
- The publisher explicitly cautions against production use without retraining because of concept drift.
- No numeric evaluation scores were extracted; the card contains only a classification-report heading without captured values.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: dima806/ai_vs_real_image_detection
Captured: Unknown. Processed: 2026-09-07T19:34:43.551913+00:00.
Checks whether the image is real or fake (AI-generated). Note to users who want to use this model in production: Beware that this model is trained on a dataset collected about 2 years ago. Since then, there is a remarkable progress in generating deepfake images with common AI tools, resulting in a significant concept drift. To mitigate that, I urge you to retrain the model using the latest available labeled data. As a quick-fix approach, simple reducing the threshold (say from default 0.5 to 0.1 or even 0.01) of labelling image as a fake may suffice. However, you will do that at your own risk, and retraining the model is the better…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014