ashishupadhyay/NSFW_DETECTION

NSFW Image Detection Model

A fine-tuned Vision Transformer (ViT) model for NSFW image classification based on the "google/vit-base-patch16-224-in21k" architecture.

Model Description

This model is a variant of the transformer encoder architecture adapted for image classification tasks. It has been fine-tuned specifically for detecting Not Safe For Work (NSFW) content in images with high accuracy.

Training Details

• Base Model: google/vit-base-patch16-224-in21k
• Image Resolution: 224x224 pixels
• Batch Size: 16
• Learning Rate: 5e-5
• Training Dataset: 80,000 images (proprietary dataset)
• Classes: 2 (normal, nsfw)

Training Stats

• Evaluation Loss: 0.07463
• Accuracy: 98.04%
• Evaluation Runtime: 304.98s
• Samples per Second: 52.46
• Steps per Second: 3.28

How to Use

Using Pipeline (Simple Method)

from PIL import Image
from transformers import pipeline

img = Image.open("<path_to_image_file>")
classifier = pipeline("image-classification", model="ashishupadhyay/NSFW_DETECTION")
classifier(img)

Using Model Directly

import torch
from PIL import Image
from transformers import AutoModelForImageClassification, ViTImageProcessor

img = Image.open("<path_to_image_file>")
model = AutoModelForImageClassification.from_pretrained("ashishupadhyay/NSFW_DETECTION")
processor = ViTImageProcessor.from_pretrained('ashishupadhyay/NSFW_DETECTION')

with torch.no_grad():
    inputs = processor(images=img, return_tensors="pt")
    outputs = model(**inputs)
    logits = outputs.logits

predicted_label = logits.argmax(-1).item()
model.config.id2label[predicted_label]

Intended Uses & Limitations

Intended Uses

• NSFW Image Classification
• Content Moderation
• Safe Content Filtering

Limitations

• Specialized for NSFW detection only
• Performance may vary for other tasks
• Should be used in conjunction with content guidelines and regulations

Important Notes

• The model is fine-tuned specifically for NSFW content detection
• Use responsibly and ethically
• Adhere to applicable content guidelines and regulations
• Performance may vary based on input image quality and characteristics

References

• Hugging Face Model Hub
• Vision Transformer (ViT) Paper
• ImageNet-21k Dataset

Disclaimer

The model's performance may be influenced by the quality and representativeness of the training data. Users should assess the model's suitability for their specific applications and datasets before implementation.