{
  "id": 121879,
  "title": "Real Vs Fake Audio Mel Frequency",
  "url": "/competitions/deepfake-detection-challenge/discussion/121879",
  "author_name": "",
  "post_date": "2019-12-16T10:15:44.479390300Z",
  "votes": 14,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Here is a comparison of Mel Frequency Diagram from Real and Fake training sample . (Fake Video and the corresponding Original ) . Is it only me ,or the Fake Audio seems to be a little blurry with little less prominent colors (decibels )</p>\n\n<p>EDIT : Thanks for all the comments and pointing out the mistake .  Also the valuable threads </p>\n\n<p><a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121861\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121861</a>\n<a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121268\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121268</a>\n<a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121694\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121694</a></p>\n\n<p>Here is the correction and as <a href=\"/cayala\">@cayala</a> pointed out I checked the audio arrays with np.array_equal() this time .</p>\n\n<p>Real (flqgmnetsg.mp4)(from set 45)\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F0e4e66da8086b6e6c3b6a3f567b5a6e0%2Freal.png?generation=1576848335307309&amp;alt=media\" alt=\"\"></p>\n\n<p>Fake (aaeflzzhvy.mp4)(from set45)</p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F567a47cd184e9c4315cb3091580ebf89%2Ffake.png?generation=1576848388484642&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": "696228",
      "postDate": "12/16/2019 10:15:44",
      "content": "<p>Here is a comparison of Mel Frequency Diagram from Real and Fake training sample . (Fake Video and the corresponding Original ) . Is it only me ,or the Fake Audio seems to be a little blurry with little less prominent colors (decibels )</p>\n\n<p>EDIT : Thanks for all the comments and pointing out the mistake .  Also the valuable threads </p>\n\n<p><a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121861\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121861</a>\n<a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121268\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121268</a>\n<a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121694\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121694</a></p>\n\n<p>Here is the correction and as <a href=\"/cayala\">@cayala</a> pointed out I checked the audio arrays with np.array_equal() this time .</p>\n\n<p>Real (flqgmnetsg.mp4)(from set 45)\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F0e4e66da8086b6e6c3b6a3f567b5a6e0%2Freal.png?generation=1576848335307309&amp;alt=media\" alt=\"\"></p>\n\n<p>Fake (aaeflzzhvy.mp4)(from set45)</p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F567a47cd184e9c4315cb3091580ebf89%2Ffake.png?generation=1576848388484642&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "Here is a comparison of Mel Frequency Diagram from Real and Fake training sample . (Fake Video and the corresponding Original ) . Is it only me ,or the Fake Audio seems to be a little blurry with little less prominent colors (decibels )\n\nEDIT : Thanks for all the comments and pointing out the mistake .  Also the valuable threads \n\nhttps://www.kaggle.com/c/deepfake-detection-challenge/discussion/121861\nhttps://www.kaggle.com/c/deepfake-detection-challenge/discussion/121268\nhttps://www.kaggle.com/c/deepfake-detection-challenge/discussion/121694\n\nHere is the correction and as @cayala pointed out I checked the audio arrays with np.array_equal() this time .\n\nReal (flqgmnetsg.mp4)(from set 45)\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F0e4e66da8086b6e6c3b6a3f567b5a6e0%2Freal.png?generation=1576848335307309&amp;alt=media)\n\n\nFake (aaeflzzhvy.mp4)(from set45)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F567a47cd184e9c4315cb3091580ebf89%2Ffake.png?generation=1576848388484642&amp;alt=media)",
      "votes": null
    },
    {
      "id": "696962",
      "postDate": "12/17/2019 09:45:19",
      "content": "<p>yep. the fake one is blurrier</p>",
      "rawMarkdown": "yep. the fake one is blurrier",
      "votes": null
    },
    {
      "id": "697413",
      "postDate": "12/17/2019 21:38:08",
      "content": "<p>Those spectograms are just the same.</p>\n\n<p>~~~\nReal (ccfoszqabv.mp4)</p>\n\n<p>array([[-2.51 , -2.133, -2.309, ..., -2.533, -2.887, -2.86 ],\n       [-2.002, -2.162, -1.846, ..., -2.95 , -2.723, -2.86 ],\n       [-1.308, -1.519, -1.733, ..., -2.695, -3.705, -3.168],\n       ...,\n       [-2.922, -3.162, -3.47 , ..., -3.39 , -3.357, -3.254],\n       [-3.11 , -3.197, -3.316, ..., -3.613, -3.404, -3.352],\n       [-3.324, -3.236, -3.256, ..., -3.502, -3.398, -3.205]],\n      dtype=float16)</p>\n\n<p>Fake (acqfdwsrhi.mp4):</p>\n\n<p>array([[-2.51 , -2.133, -2.309, ..., -2.533, -2.887, -2.86 ],\n       [-2.002, -2.162, -1.846, ..., -2.95 , -2.723, -2.86 ],\n       [-1.308, -1.519, -1.733, ..., -2.695, -3.705, -3.168],\n       ...,\n       [-2.922, -3.162, -3.47 , ..., -3.39 , -3.357, -3.254],\n       [-3.11 , -3.197, -3.316, ..., -3.613, -3.404, -3.352],\n       [-3.324, -3.236, -3.256, ..., -3.502, -3.398, -3.205]],\n      dtype=float16)</p>\n\n<p>Diff</p>\n\n<p>np.sum(np.abs(r_mel_spec - f_mel_spec)) = 0.0\n ~~~</p>\n\n<p>I reckon the issue with this video is her face, that has been swaped.</p>",
      "rawMarkdown": "Those spectograms are just the same.\n\n~~~\nReal (ccfoszqabv.mp4)\n\narray([[-2.51 , -2.133, -2.309, ..., -2.533, -2.887, -2.86 ],\n       [-2.002, -2.162, -1.846, ..., -2.95 , -2.723, -2.86 ],\n       [-1.308, -1.519, -1.733, ..., -2.695, -3.705, -3.168],\n       ...,\n       [-2.922, -3.162, -3.47 , ..., -3.39 , -3.357, -3.254],\n       [-3.11 , -3.197, -3.316, ..., -3.613, -3.404, -3.352],\n       [-3.324, -3.236, -3.256, ..., -3.502, -3.398, -3.205]],\n      dtype=float16)\n\nFake (acqfdwsrhi.mp4):\n\narray([[-2.51 , -2.133, -2.309, ..., -2.533, -2.887, -2.86 ],\n       [-2.002, -2.162, -1.846, ..., -2.95 , -2.723, -2.86 ],\n       [-1.308, -1.519, -1.733, ..., -2.695, -3.705, -3.168],\n       ...,\n       [-2.922, -3.162, -3.47 , ..., -3.39 , -3.357, -3.254],\n       [-3.11 , -3.197, -3.316, ..., -3.613, -3.404, -3.352],\n       [-3.324, -3.236, -3.256, ..., -3.502, -3.398, -3.205]],\n      dtype=float16)\n\nDiff\n\nnp.sum(np.abs(r_mel_spec - f_mel_spec)) = 0.0\n ~~~\n\nI reckon the issue with this video is her face, that has been swaped.",
      "votes": null
    },
    {
      "id": "699458",
      "postDate": "12/20/2019 13:27:45",
      "content": "<p>Thanks so much :) . I realized it later . Updated . This time its truly different.</p>",
      "rawMarkdown": "Thanks so much :) . I realized it later . Updated . This time its truly different.",
      "votes": null
    },
    {
      "id": "699487",
      "postDate": "12/20/2019 14:08:35",
      "content": "<p>updated. Earlier one was affected by the syndrome \"You see, what you want to see !\"</p>",
      "rawMarkdown": "updated. Earlier one was affected by the syndrome \"You see, what you want to see !\"",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 696962,
      "author_name": "moewie94",
      "author_url": "",
      "post_date": "12/17/2019 09:45:19",
      "content": "<p>yep. the fake one is blurrier</p>",
      "votes": null,
      "replies": [
        {
          "id": 699487,
          "author_name": "phoenix9032",
          "author_url": "",
          "post_date": "12/20/2019 14:08:35",
          "content": "<p>updated. Earlier one was affected by the syndrome \"You see, what you want to see !\"</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 697413,
      "author_name": "cayala",
      "author_url": "",
      "post_date": "12/17/2019 21:38:08",
      "content": "<p>Those spectograms are just the same.</p>\n\n<p>~~~\nReal (ccfoszqabv.mp4)</p>\n\n<p>array([[-2.51 , -2.133, -2.309, ..., -2.533, -2.887, -2.86 ],\n       [-2.002, -2.162, -1.846, ..., -2.95 , -2.723, -2.86 ],\n       [-1.308, -1.519, -1.733, ..., -2.695, -3.705, -3.168],\n       ...,\n       [-2.922, -3.162, -3.47 , ..., -3.39 , -3.357, -3.254],\n       [-3.11 , -3.197, -3.316, ..., -3.613, -3.404, -3.352],\n       [-3.324, -3.236, -3.256, ..., -3.502, -3.398, -3.205]],\n      dtype=float16)</p>\n\n<p>Fake (acqfdwsrhi.mp4):</p>\n\n<p>array([[-2.51 , -2.133, -2.309, ..., -2.533, -2.887, -2.86 ],\n       [-2.002, -2.162, -1.846, ..., -2.95 , -2.723, -2.86 ],\n       [-1.308, -1.519, -1.733, ..., -2.695, -3.705, -3.168],\n       ...,\n       [-2.922, -3.162, -3.47 , ..., -3.39 , -3.357, -3.254],\n       [-3.11 , -3.197, -3.316, ..., -3.613, -3.404, -3.352],\n       [-3.324, -3.236, -3.256, ..., -3.502, -3.398, -3.205]],\n      dtype=float16)</p>\n\n<p>Diff</p>\n\n<p>np.sum(np.abs(r_mel_spec - f_mel_spec)) = 0.0\n ~~~</p>\n\n<p>I reckon the issue with this video is her face, that has been swaped.</p>",
      "votes": null,
      "replies": [
        {
          "id": 699458,
          "author_name": "phoenix9032",
          "author_url": "",
          "post_date": "12/20/2019 13:27:45",
          "content": "<p>Thanks so much :) . I realized it later . Updated . This time its truly different.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "696228": "Here is a comparison of Mel Frequency Diagram from Real and Fake training sample . (Fake Video and the corresponding Original ) . Is it only me ,or the Fake Audio seems to be a little blurry with little less prominent colors (decibels )\n\nEDIT : Thanks for all the comments and pointing out the mistake .  Also the valuable threads \n\nhttps://www.kaggle.com/c/deepfake-detection-challenge/discussion/121861\nhttps://www.kaggle.com/c/deepfake-detection-challenge/discussion/121268\nhttps://www.kaggle.com/c/deepfake-detection-challenge/discussion/121694\n\nHere is the correction and as @cayala pointed out I checked the audio arrays with np.array_equal() this time .\n\nReal (flqgmnetsg.mp4)(from set 45)\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F0e4e66da8086b6e6c3b6a3f567b5a6e0%2Freal.png?generation=1576848335307309&amp;alt=media)\n\n\nFake (aaeflzzhvy.mp4)(from set45)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F567a47cd184e9c4315cb3091580ebf89%2Ffake.png?generation=1576848388484642&amp;alt=media)",
    "696962": "yep. the fake one is blurrier",
    "697413": "Those spectograms are just the same.\n\n~~~\nReal (ccfoszqabv.mp4)\n\narray([[-2.51 , -2.133, -2.309, ..., -2.533, -2.887, -2.86 ],\n       [-2.002, -2.162, -1.846, ..., -2.95 , -2.723, -2.86 ],\n       [-1.308, -1.519, -1.733, ..., -2.695, -3.705, -3.168],\n       ...,\n       [-2.922, -3.162, -3.47 , ..., -3.39 , -3.357, -3.254],\n       [-3.11 , -3.197, -3.316, ..., -3.613, -3.404, -3.352],\n       [-3.324, -3.236, -3.256, ..., -3.502, -3.398, -3.205]],\n      dtype=float16)\n\nFake (acqfdwsrhi.mp4):\n\narray([[-2.51 , -2.133, -2.309, ..., -2.533, -2.887, -2.86 ],\n       [-2.002, -2.162, -1.846, ..., -2.95 , -2.723, -2.86 ],\n       [-1.308, -1.519, -1.733, ..., -2.695, -3.705, -3.168],\n       ...,\n       [-2.922, -3.162, -3.47 , ..., -3.39 , -3.357, -3.254],\n       [-3.11 , -3.197, -3.316, ..., -3.613, -3.404, -3.352],\n       [-3.324, -3.236, -3.256, ..., -3.502, -3.398, -3.205]],\n      dtype=float16)\n\nDiff\n\nnp.sum(np.abs(r_mel_spec - f_mel_spec)) = 0.0\n ~~~\n\nI reckon the issue with this video is her face, that has been swaped.",
    "699458": "Thanks so much :) . I realized it later . Updated . This time its truly different.",
    "699487": "updated. Earlier one was affected by the syndrome \"You see, what you want to see !\""
  },
  "source": "meta"
}