nsfwspy / nsfwspy.js Goto Github PK

const NSFWNET_WEIGHTS_PATH ='models/model.json';

const IMAGE_SIZE = 224;
const IMAGE_CROP_SIZE = 224;
const TOPK_PREDICTIONS = 4;

const NSFW_CLASSES = {
  0: 'Hentai',
  1: 'Neural',
  2: 'Porn',
  3: 'Sexy',
};


let nsfwnet;
const nsfwnetDemo = async () => {

  nsfwnet = await tf.loadGraphModel(NSFWNET_WEIGHTS_PATH);

  nsfwnet.predict(tf.zeros([1, IMAGE_CROP_SIZE, IMAGE_CROP_SIZE, 3])).dispose();

  console.log('Model Warm complete');

  const image_Element = document.getElementById('test_draw');
  if (image_Element.complete && image_Element.naturalHeight !== 0) {

    predict(image_Element);
    image_Element.style.display = '';
  } 
  document.getElementById('file-container').style.display = '';
};


async function predict(imgElement) {
  

  const logits = tf.tidy(() => {

    const img = tf.browser.fromPixels(imgElement).toFloat();
    const crop_image = tf.slice(img, [16, 16, 0], [224, 224, -1]);
    const img_reshape = tf.reverse(crop_image, [-1]);

    let imagenet_mean = tf.expandDims([103.94, 116.78, 123.68], 0);
    imagenet_mean = tf.expandDims(imagenet_mean, 0);

    const normalized = img_reshape.sub(imagenet_mean);

    const batched = normalized.reshape([1, IMAGE_CROP_SIZE, IMAGE_CROP_SIZE, 3]);

    return nsfwnet.predict(batched);
  });

  const classes = await getTopKClasses(logits, TOPK_PREDICTIONS);

  display(classes);
}


async function getTopKClasses(logits, topK){
  const values = await logits.data();
  sortArray = Array.from(values).map((value, index) => {
    return {
      value: value,
      index: index
    };
  }
  ).sort((a, b) => {
    return b.value - a.value;
  }).slice(0, topK);

  return sortArray.map(x => {
    return {
      className: NSFW_CLASSES[x.index],
      probability: x.value
    };
  }
  );
}


function display(classes){
  console.log(classes);
}



nsfwnetDemo();

$ node .
2023-07-09 16:20:01.450460: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
\node_modules\@tensorflow\tfjs-converter\dist\tf-converter.node.js:30397
        this.version = graph.versions.producer + "." + graph.versions.minConsumer;
                                      ^

TypeError: Cannot read properties of undefined (reading 'producer')
    at GraphModel.loadSync (\node_modules\@tensorflow\tfjs-converter\dist\tf-converter.node.js:30397:39)
    at \node_modules\@tensorflow\tfjs-converter\dist\tf-converter.node.js:30373:72

Node.js v18.14.2

const nsfwSpy = new NsfwSpy("../models/mobilenet-v1.0.0/model.json");

const filePath = "/home/fmx/Pictures/Screenshot_20230528_120040.png";
await nsfwSpy.load();
const result = await nsfwSpy.classifyImageFile(filePath);

console.log("Console Log!!")
console.log(result);```