Guard-Against-Unsafe-Content-Siglip2
Guard-Against-Unsafe-Content-Siglip2 is an image classification vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for a single-label classification task. It is designed to detect NSFW content, including vulgarity and nudity, using the SiglipForImageClassification architecture.
[!Important] Accuracy : 99 [ nudity, vulgarity, explicit content]
The model categorizes images into two classes:
- Class 0: "Unsafe Content" – indicating that the image contains vulgarity, nudity, or explicit content.
- Class 1: "Safe Content" – indicating that the image is appropriate and does not contain any unsafe elements.
Run with Transformers🤗
!pip install -q transformers torch pillow gradio
import gradio as gr
from transformers import AutoImageProcessor
from transformers import SiglipForImageClassification
from transformers.image_utils import load_image
from PIL import Image
import torch
# Load model and processor
model_name = "prithivMLmods/Guard-Against-Unsafe-Content-Siglip2"
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)
def nsfw_detection(image):
"""Predicts NSFW probability scores for an 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()
labels = model.config.id2label
predictions = {labels[i]: round(probs[i], 3) for i in range(len(probs))}
return predictions
# Create Gradio interface
iface = gr.Interface(
fn=nsfw_detection,
inputs=gr.Image(type="numpy"),
outputs=gr.Label(label="NSFW Content Detection"),
title="NSFW Image Detection",
description="Upload an image to check if it contains unsafe content such as vulgarity or nudity."
)
# Launch the app
if __name__ == "__main__":
iface.launch()
TrainOutput(global_step=376, training_loss=0.11756020403922872, metrics={'train_runtime': 597.6963, 'train_samples_per_second': 20.077, 'train_steps_per_second': 0.629, 'total_flos': 1.005065949855744e+18, 'train_loss': 0.11756020403922872, 'epoch': 2.0})
Intended Use:
The Guard-Against-Unsafe-Content-Siglip2 model is designed to detect inappropriate and explicit content in images. It helps distinguish between safe and unsafe images based on the presence of vulgarity, nudity, or other NSFW elements.
Potential Use Cases:
- NSFW Content Detection: Identifying images containing explicit content to help filter inappropriate material.
- Content Moderation: Assisting platforms in filtering out unsafe images before they are shared publicly.
- Parental Controls: Enabling automated filtering of explicit images in child-friendly environments.
- Safe Image Classification: Helping AI-powered applications distinguish between safe and unsafe content for appropriate usage.