Erax NSFW V1.0

🚀 EraX-NSFW-V1.0

EraX-NSFW-V1.0是一款高效的不适宜内容（NSFW）检测模型，在发布前的图像和视频管控，以及限制儿童接触有害出版物方面表现出色。你既可以预测类别及其边界框，还能对预测出的有害物体进行打码处理，甚至对整个图像进行打码。

🚀 快速开始

安装必要的包

pip install ultralytics supervision huggingface-hub

下载预训练模型

from huggingface_hub import snapshot_download
snapshot_download(repo_id="erax-ai/EraX-NSFW-V1.0", local_dir="./", force_download=True)

简单用例

from ultralytics import YOLO
from PIL import Image
import supervision as sv
import numpy as np

IOU_THRESHOLD        = 0.3
CONFIDENCE_THRESHOLD = 0.2

# pretrained_path = "erax_nsfw_yolo11n.pt"
# pretrained_path = "erax_nsfw_yolo11s.pt"
pretrained_path = "erax_nsfw_yolo11m.pt"

image_path_list = ["test_images/img_1.jpg", "test_images/img_2.jpg"]

model = YOLO(pretrained_path)
results = model(image_path_list,
                  conf=CONFIDENCE_THRESHOLD,
                  iou=IOU_THRESHOLD
                )


for result in results:
    annotated_image = result.orig_img.copy()
    h, w = annotated_image.shape[:2]
    anchor = h if h > w else w

    detections = sv.Detections.from_ultralytics(result)
    label_annotator = sv.LabelAnnotator(text_color=sv.Color.BLACK,
                                        text_position=sv.Position.CENTER,
                                        text_scale=anchor/1700)
    
    pixelate_annotator = sv.PixelateAnnotator(pixel_size=anchor/50)
    
    annotated_image = pixelate_annotator.annotate(
        scene=annotated_image.copy(),
        detections=detections
    )


    annotated_image = label_annotator.annotate(
        annotated_image,
        detections=detections
    )

    
    sv.plot_image(annotated_image, size=(10, 10))

✨ 主要特性

• 可预测类别及其边界框。
• 能对预测出的有害物体进行打码处理，甚至对整个图像进行打码。

📦 安装指南

安装必要的包：

pip install ultralytics supervision huggingface-hub

💻 使用示例

基础用法

from ultralytics import YOLO
from PIL import Image
import supervision as sv
import numpy as np

IOU_THRESHOLD        = 0.3
CONFIDENCE_THRESHOLD = 0.2

# pretrained_path = "erax_nsfw_yolo11n.pt"
# pretrained_path = "erax_nsfw_yolo11s.pt"
pretrained_path = "erax_nsfw_yolo11m.pt"

image_path_list = ["test_images/img_1.jpg", "test_images/img_2.jpg"]

model = YOLO(pretrained_path)
results = model(image_path_list,
                  conf=CONFIDENCE_THRESHOLD,
                  iou=IOU_THRESHOLD
                )


for result in results:
    annotated_image = result.orig_img.copy()
    h, w = annotated_image.shape[:2]
    anchor = h if h > w else w

    detections = sv.Detections.from_ultralytics(result)
    label_annotator = sv.LabelAnnotator(text_color=sv.Color.BLACK,
                                        text_position=sv.Position.CENTER,
                                        text_scale=anchor/1700)
    
    pixelate_annotator = sv.PixelateAnnotator(pixel_size=anchor/50)
    
    annotated_image = pixelate_annotator.annotate(
        scene=annotated_image.copy(),
        detections=detections
    )


    annotated_image = label_annotator.annotate(
        annotated_image,
        detections=detections
    )

    
    sv.plot_image(annotated_image, size=(10, 10))

📚 详细文档

模型详情 / 概述

标签

训练配置

• 模型权重文件：
  • Nano：erax_nsfw_yolo11n.pt
  • Small：erax_nsfw_yolo11s.pt
  • Medium：erax_nsfw_yolo11m.pt
• 训练轮数：100
• 学习率：0.01
• 批次大小：208
• 图像尺寸：640x640
• 训练服务器：8 x NVIDIA RTX A4000（16GB GDDR6）
• 训练时间：约10小时

训练验证结果

训练和验证损失

混淆矩阵

更多示例

1. 示例01：

2. 示例02：

3. 示例03：使用性行为类别最为安全，因为它会覆盖整个上下文。 无性行为类别 | 有性行为类别 :-------------------------:|:-------------------------: | | |

训练

训练脚本：https://github.com/EraX-JS-Company/EraX-NSFW-V1.0

🔧 技术细节

模型基于YOLO11架构进行训练，使用私有数据集（来自互联网），训练集包含31890张图像，验证集包含11538张图像。训练过程中，设置了100个训练轮数，学习率为0.01，批次大小为208，图像尺寸为640x640，使用8 x NVIDIA RTX A4000（16GB GDDR6）服务器进行训练，训练时间约为10小时。

📄 许可证

本项目采用Apache 2.0许可证。

引用

如果您觉得我们的项目有用，请给我们的仓库点个星，并按以下方式引用我们的工作：

@article{EraX-NSFW-V1.0,
  author    = {Phạm Đình Thục and
              Mr. Nguyễn Anh Nguyên and
              Đoàn Thành Khang and
              Mr. Trần Hải Khương and
              Mr. Trương Công Đức and 
              Phan Nguyễn Tuấn Kha and 
              Phạm Huỳnh Nhật},
  title     = {EraX-NSFW-V1.0: A Highly Efficient Model for NSFW Detection},
  organization={EraX JS Company},
  year={2024},
  url={https://huggingface.co/erax-ai/EraX-NSFW-V1.0}
}

⚠️ 重要提示

此媒体包含一些人可能会觉得冒犯的敏感内容（如裸露、暴力、亵渎、色情内容）。您必须年满18岁才能查看此内容。