File size: 5,447 Bytes
ba864ff 84d6f0b d070f62 87b9bf7 3daf43e 6ec6a8e 87b9bf7 72d84bb 917cd76 87b9bf7 6ec6a8e d070f62 6ec6a8e 917cd76 6ec6a8e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 |
---
license: apache-2.0
datasets:
- strangerguardhf/NSFW-MultiDomain-Classification-v2.0
language:
- en
base_model:
- google/vit-base-patch16-224
pipeline_tag: image-classification
library_name: transformers
tags:
- explicit-content-detection
- mini
- art
- sensual-content-detection
- Anime
---
# **vit-mini-explicit-content**
> **vit-mini-explicit-content** is an image classification vision-language model fine-tuned from **vit-base-patch16-224-in21k** for a single-label classification task. It categorizes images based on their explicitness using the **ViTForImageClassification** architecture.
> \[!Note]
> This model is designed to promote safe, respectful, and responsible online spaces. It does **not** generate explicit content; it only classifies images. Misuse may violate platform or regional policies and is strongly discouraged.
> [!Note]
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale : https://arxiv.org/abs/2010.11929, Visual Transformers: Token-based Image Representation and Processing for Computer Vision: https://arxiv.org/pdf/2006.03677
> [!Important]
Note: Explicit, sensual, and pornographic content may appear in the results; however, all of them are considered not safe for work.
```py
Classification Report:
precision recall f1-score support
Anime Picture 0.9077 0.7937 0.8469 5600
Extincing & Sensual 0.9245 0.9717 0.9475 5618
Hentai 0.8680 0.9391 0.9021 5600
Pornography 0.9614 0.9544 0.9579 5970
Safe for Work 0.9235 0.9235 0.9235 6000
accuracy 0.9171 28788
macro avg 0.9170 0.9165 0.9156 28788
weighted avg 0.9177 0.9171 0.9163 28788
```
---
The model categorizes images into five classes:
* **Class 0:** Anime Picture
* **Class 1:** Enticing & Sensual
* **Class 2:** Hentai
* **Class 3:** Pornography
* **Class 4:** Safe for Work
# **Run with Transformers**
```python
!pip install -q transformers torch pillow gradio
```
```python
import gradio as gr
from transformers import ViTImageProcessor, ViTForImageClassification
from PIL import Image
import torch
# Load model and processor
model_name = "prithivMLmods/vit-mini-explicit-content" # Updated model path
model = ViTForImageClassification.from_pretrained(model_name)
processor = ViTImageProcessor.from_pretrained(model_name)
# Updated label mapping
labels = {
"0": "Anime Picture",
"1": "Enticing & Sensual",
"2": "Hentai",
"3": "Pornography",
"4": "Safe for Work"
}
def explicit_content_detection(image):
"""Predicts the type of content in the image."""
image = Image.fromarray(image).convert("RGB")
inputs = processor(images=image, return_tensors="pt")
with torch.no_grad():
outputs = model(**inputs)
logits = outputs.logits
probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist()
predictions = {labels[str(i)]: round(probs[i], 3) for i in range(len(probs))}
return predictions
# Create Gradio interface
iface = gr.Interface(
fn=explicit_content_detection,
inputs=gr.Image(type="numpy"),
outputs=gr.Label(label="Prediction Scores"),
title="vit-mini-explicit-content",
description="Upload an image to classify whether it is anime, enticing & sensual, hentai, pornographic, or safe for work."
)
# Launch the app
if __name__ == "__main__":
iface.launch()
```
---
## Demo Inference
> [!warning]
Anime Picture
> [!warning]
Extincing & Sensual
> [!warning]
Hentai
> [!warning]
Pornography
> [!warning]
Safe for Work
---
# **Recommended Use Cases**
* Image moderation pipelines
* Parental and institutional content filters
* Dataset cleansing before training
* Online safety and well-being platforms
* Enhancing search engine filtering
# **Discouraged / Prohibited Use**
* Non-consensual or malicious monitoring
* Automated judgments without human review
* Misrepresentation of moderation systems
* Use in unlawful or unethical surveillance
* Harassment, exploitation, or shaming |