|
--- |
|
license: apache-2.0 |
|
base_model: microsoft/swin-tiny-patch4-window7-224 |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- imagefolder |
|
metrics: |
|
- accuracy |
|
- precision |
|
- recall |
|
- f1 |
|
model-index: |
|
- name: FaceAIorNot |
|
results: |
|
- task: |
|
name: Image Classification |
|
type: image-classification |
|
dataset: |
|
name: imagefolder |
|
type: imagefolder |
|
config: default |
|
split: train |
|
args: default |
|
metrics: |
|
- name: Accuracy |
|
type: accuracy |
|
value: 0.9935440994968195 |
|
- name: Precision |
|
type: precision |
|
value: 0.9925121677274429 |
|
- name: Recall |
|
type: recall |
|
value: 0.9947467166979362 |
|
- name: F1 |
|
type: f1 |
|
value: 0.9936281859070465 |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
# FaceAIorNot |
|
|
|
Face AI or Not |
|
|
|
This model is a fine-tuned version of [microsoft/swin-tiny-patch4-window7-224](https://huggingface.co/microsoft/swin-tiny-patch4-window7-224) on the imagefolder dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.0233 |
|
- Accuracy: 0.9935 |
|
- Precision: 0.9925 |
|
- Recall: 0.9947 |
|
- F1: 0.9936 |
|
|
|
## Model description |
|
|
|
Two classes: AI-generated, Not AI-generated |
|
|
|
## Intended uses & limitations |
|
|
|
Classify an face image if is generated by AI. |
|
The classify result may not is 100% right. |
|
|
|
## Training and evaluation data |
|
|
|
Finetune in 105,330 face images. |
|
17 datasets. |
|
14 AI Image Generation Techniques. |
|
50% real faces and 50% AI-generated faces. |
|
Data set cut into 90% Train set, 10% Test set(evaluation set). |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 5e-05 |
|
- train_batch_size: 32 |
|
- eval_batch_size: 32 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 4 |
|
- total_train_batch_size: 128 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- lr_scheduler_warmup_ratio: 0.1 |
|
- num_epochs: 5 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 | |
|
|:-------------:|:-----:|:----:|:---------------:|:--------:|:---------:|:------:|:------:| |
|
| 0.0862 | 1.0 | 740 | 0.0694 | 0.9740 | 0.9731 | 0.9756 | 0.9743 | |
|
| 0.0914 | 2.0 | 1481 | 0.0396 | 0.9862 | 0.9814 | 0.9916 | 0.9865 | |
|
| 0.0184 | 3.0 | 2222 | 0.0784 | 0.9777 | 0.9618 | 0.9955 | 0.9783 | |
|
| 0.0181 | 4.0 | 2963 | 0.0330 | 0.9907 | 0.9879 | 0.9938 | 0.9908 | |
|
| 0.03 | 4.99 | 3700 | 0.0233 | 0.9935 | 0.9925 | 0.9947 | 0.9936 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.34.0 |
|
- Pytorch 2.1.1+cu118 |
|
- Datasets 2.14.5 |
|
- Tokenizers 0.14.1 |
|
|
