基于 NSFW Model 图片识别鉴黄后面更新视频检测

识别效果

在这里插入图片描述

在这里插入图片描述
推荐SFW 0.98 色情 NSFW 0.015

在这里插入图片描述

推荐SFW 0.00 色情 NSFW 0.99 为色情图片

中文github代码下载 - 原文地址

环境条件

所有代码都应该与Python 3.6and兼容Tensorflow 1.x（用 1.12 测试）。模型实现可以在中找到model.py

作者python环境

Python 3.6.13 :: Anaconda, Inc.

pip list

absl-py==1.1.0
astor==0.8.1
cached-property==1.5.2
certifi==2021.5.30
cycler==0.11.0
dataclasses==0.8
decorator==4.4.2
gast==0.5.3
grpcio==1.46.3
h5py==3.1.0
imageio==2.15.0
importlib-metadata==4.8.3
Keras-Applications==1.0.8
Keras-Preprocessing==1.1.2
kiwisolver==1.3.1
Markdown==3.3.7
matplotlib==3.3.4
networkx==2.5.1
numpy==1.16.2
Pillow==8.4.0
protobuf==4.21.0
pyparsing==3.0.9
python-dateutil==2.8.2
PyWavelets==1.1.1
scikit-image==0.17.2
scipy==1.5.4
six==1.16.0
tensorboard==1.12.2
tensorflow==1.12.0
termcolor==1.1.0
tifffile==2020.9.3
typing_extensions==4.1.1
Werkzeug==2.0.3
wincertstore==0.2
zipp==3.6.0

安装依赖

方法 1.
pip install tensorflow== 1.12
pip installnumpy==1.16.2
pip install scikit-image==0.17.2方法 2.
pip install -r requirements.txt

使用方法

这里注意一下只支持.jpg类型图片

python classify_nsfw.py -m data/open_nsfw-weights.npy data/test.jpg

SFW score推荐比例  , NSFW score 不推荐比例
Results for 'test.jpg'SFW score:	0.9355766177177429 NSFW score:	0.06442338228225708

视频或者图片批量检测

python python http_nsfw.py	POST请求 http://127.0.0.1:2200/?token=aid.sd5@ns85
///图片格式///
{"temp_path":"C:/Users/lijie/Desktop/新-短视频/视频分割检测/66/","type":"image","valve":0.8,"src":["http://aliyuncs.com/image/2022-06/16/I1nfUPElFH6oz8ux.jpeg","http://aliyuncs.com/image/2022-06/16/OGCVPdDnV1wLdTn9.png"]
}
//视频格式//
{"temp_path":"C:/Users/lijie/Desktop/新-短视频/视频分割检测/66/","type":"video","valve":0.8,"src":"http://aliyuncs.com/image/2022-06/16/OGCVPd.mp4"
}返回信息 
{"isViolation": false, //是否涩情"tNsfw": "0.07921311259269714","tsfw": "0.9207869172096252"
}

##http_nsfw.py 代码

#!/usr/bin/env python
import sys
import argparse
import tensorflow as tffrom model import OpenNsfwModel, InputType
from image_utils import create_tensorflow_image_loader
from image_utils import create_yahoo_image_loader
from pathlib import Path
import subprocess
import os
import numpy as npfrom flask import request
from flask import jsonify
from flask import Flask
import json
from flask import jsonify
import urllib.requestclass main(object):def __init__(self):self.model =''self.sess =''def test(self):print(666)def model_load(self, weights=""):model = OpenNsfwModel()#创建加载NSF模型input_type = InputType['TENSOR'] #模型类型 [InputType.TENSOR.name.lower(), InputType.BASE64_JPEG.name.lower()])model.build(weights_path=weights, input_type=input_type)self.model = model;print('模型加载OK')self.sess =tf.Session() #创建对象self.sess.run(tf.global_variables_initializer())def distinguish(self,input_file):fn_load_image = Nonefn_load_image = create_yahoo_image_loader() #创建雅虎图片加载器image = fn_load_image(input_file)predictions = \self.sess.run(self.model.predictions,feed_dict={self.model.input: image}) #识别# print("Results for '{}'".format(args.input_file))# print("\tSFW score:\t{}\n\tNSFW score:\t{}".format(*predictions[0]))return {"tSFW":predictions[0][0],"tNSFW":predictions[0][1]}# if __name__ == "__main__":def save_image(video_path, image_dir, fps):#文件夹创建if not os.path.exists(image_dir):os.makedirs(image_dir)#-vf scale=320:240cmd_str = f'ffmpeg -i {video_path}  -f image2 -r {fps} {image_dir}/%d.jpeg -loglevel quiet 'subprocess.run(cmd_str, encoding="utf-8" , shell=True)app = Flask(__name__)@app.route('/', methods=['POST'])
def distinguish():token = request.values.get("token")if token != 'aid.sd5@ns85': #tokenreturn '403'params = json.loads(request.data)try:# 不能确定正确执行的代码params = json.loads(request.data) # 将json字符串转为dictexcept:return '405 参数错误'distinguish = '';is_violation = False;#视频识别if params['type'] == 'video':#视频进行抽针处理save_image(params['src'],params['temp_path'],1)#保存帧数1# os.listdir()方法获取文件夹名字，返回数组file_name_list = os.listdir(params['temp_path'])for index,img in enumerate(file_name_list):img_str = params['temp_path'] + imgdistinguish =  Model.distinguish(img_str)if distinguish['tNSFW'] > params['valve']:is_violation = TruebreakreturnData =  {"tsfw":f"{distinguish['tSFW']}","tNsfw":f"{distinguish['tNSFW']}","isViolation":is_violation}return jsonify(returnData);#图片识别if params['type'] == 'image':#文件夹创建if not os.path.exists(params['temp_path']):os.makedirs(params['temp_path'])n = 0#用来计数或命名for img in params['src']:n += 1img_url = params['temp_path']+"%s.jpeg" % nurllib.request.urlretrieve(img, img_url)#保存图片到本地distinguish = Model.distinguish(img_url)if distinguish['tNSFW'] > params['valve']:is_violation = TruebreakreturnData =  {"tsfw":f"{distinguish['tSFW']}","tNsfw":f"{distinguish['tNSFW']}","isViolation":is_violation}return jsonify(returnData);return "Ok"if __name__ == '__main__':Model =  main()Model.model_load("data/open_nsfw-weights.npy")app.run(host='0.0.0.0', port=2200, debug=False)

出现错误

caffe: TypeError: _open() got an unexpected keyword argument ‘as_grey‘“as_grey” 实际应当修改为 “as_gray”，原因是 scikit-image 的 0.17.2 版本修改了参数名称ValueError: Object arrays cannot be loaded when allow_pickle=False
自Numpy 1.16.3版本发行之后，函数 numpy.load() 和 numpy.lib.format.read_array() 采用allow_pickle关键字，现在默认为False以响
pip install numpy=1.16.2

Note: Currently only jpeg images are supported.

classify_nsfw.py accepts some optional parameters you may want to play around with:

usage: classify_nsfw.py [-h] -m MODEL_WEIGHTS [-l {yahoo,tensorflow}][-t {tensor,base64_jpeg}]input_jpeg_filepositional arguments:input_file            Path to the input image. Only jpeg images aresupported.optional arguments:-h, --help            show this help message and exit-m MODEL_WEIGHTS, --model_weights MODEL_WEIGHTSPath to trained model weights file-l {yahoo,tensorflow}, --image_loader {yahoo,tensorflow}image loading mechanism-i {tensor,base64_jpeg}, --input_type {tensor,base64_jpeg}input type

-l/–image-loader

The classification tool supports two different image loading mechanisms.

yahoo (default) replicates yahoo’s original image loading and preprocessing. Use this option if you want the same
results as with the original implementation
tensorflow is an image loader which uses tensorflow exclusively (no dependencies on PIL, skimage, etc.). Tries
to replicate the image loading mechanism used by the original caffe implementation, differs a bit though due to
different jpeg and resizing implementations.
See this issue for details.

Note: Classification results may vary depending on the selected image loader!

-i/–input_type

Determines if the model internally uses a float tensor (tensor - [None, 224, 224, 3] - default) or a base64 encoded
string tensor (base64_jpeg - [None, ]) as input. If base64_jpeg is used, then the tensorflow image loader will
be used, regardless of the -l/–image-loader argument.

Tools

The tools folder contains some utility scripts to test the model.

create_predict_request.py

Takes an input image and generates a json file suitable for prediction requests to a Open NSFW Model deployed
with Google Cloud ML Engine (gcloud ml-engine predict)
or tensorflow-serving.

export_savedmodel.py

Exports the model using the tensorflow serving export api (SavedModel). The export can be used to deploy the model
on Google Cloud ML Engine
, Tensorflow Serving or on mobile (haven’t tried that one yet).

export_tflite.py

Exports the model in TFLite format. Use this one if you want to run inference on
mobile or IoT devices. Please note that the base64_jpeg input type does not work with TFLite since the standard
runtime lacks a number of required tensorflow operations.

export_graph.py

Exports the tensorflow graph and checkpoint. Freezes and optimizes the graph per default for improved inference and
deployment usage (e.g. Android, iOS, etc.). Import the graph with tf.import_graph_def.