{
  "id": 130635,
  "title": "face detection method for inference",
  "url": "/competitions/deepfake-detection-challenge/discussion/130635",
  "author_name": "",
  "post_date": "2020-02-15T12:22:09.560640100Z",
  "votes": 2,
  "comment_count": 4,
  "views": 0,
  "content": "<p>How are you dealing with the time constraints of submission?  and what's your priority accuracy of faces or more frames for prediction\nIn my case, It takes almost 16s to detect face in all the 300 frames of a video(including video read time). Which is not good enough for submission. Currently, I've limited myself to 8-24 frames for each video in submission.\nBTW I'm using OpenCV face detection with cnn model.</p>",
  "messages": [
    {
      "id": "746696",
      "postDate": "02/15/2020 12:22:09",
      "content": "<p>How are you dealing with the time constraints of submission?  and what's your priority accuracy of faces or more frames for prediction\nIn my case, It takes almost 16s to detect face in all the 300 frames of a video(including video read time). Which is not good enough for submission. Currently, I've limited myself to 8-24 frames for each video in submission.\nBTW I'm using OpenCV face detection with cnn model.</p>",
      "rawMarkdown": "How are you dealing with the time constraints of submission?  and what's your priority accuracy of faces or more frames for prediction\nIn my case, It takes almost 16s to detect face in all the 300 frames of a video(including video read time). Which is not good enough for submission. Currently, I've limited myself to 8-24 frames for each video in submission.\nBTW I'm using OpenCV face detection with cnn model.",
      "votes": null
    },
    {
      "id": "751934",
      "postDate": "02/20/2020 16:50:23",
      "content": "<p>Are you even allowed to use openCV? </p>\n\n<p>In the code requirements is stated: No custom packages enabled in your submission notebook</p>",
      "rawMarkdown": "Are you even allowed to use openCV? \n\nIn the code requirements is stated: No custom packages enabled in your submission notebook",
      "votes": null
    },
    {
      "id": "751940",
      "postDate": "02/20/2020 16:54:48",
      "content": "<p>This is why I use BlazeFace, it's not perfect but it's fast. <a href=\"https://www.kaggle.com/humananalog/starter-blazeface-pytorch\">https://www.kaggle.com/humananalog/starter-blazeface-pytorch</a></p>",
      "rawMarkdown": "This is why I use BlazeFace, it's not perfect but it's fast. https://www.kaggle.com/humananalog/starter-blazeface-pytorch",
      "votes": null
    },
    {
      "id": "751969",
      "postDate": "02/20/2020 17:26:11",
      "content": "<p>Thank you for your suggestion. I'll definitely look into it.</p>",
      "rawMarkdown": "Thank you for your suggestion. I'll definitely look into it.",
      "votes": null
    },
    {
      "id": "751970",
      "postDate": "02/20/2020 17:27:23",
      "content": "<p>As far as I know, they mean you can't install them using the standard method of installing custom packages in Kaggle notebook. You'll have to manually load them through the dataset.</p>",
      "rawMarkdown": "As far as I know, they mean you can't install them using the standard method of installing custom packages in Kaggle notebook. You'll have to manually load them through the dataset.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 751934,
      "author_name": "gmm297",
      "author_url": "",
      "post_date": "02/20/2020 16:50:23",
      "content": "<p>Are you even allowed to use openCV? </p>\n\n<p>In the code requirements is stated: No custom packages enabled in your submission notebook</p>",
      "votes": null,
      "replies": [
        {
          "id": 751970,
          "author_name": "ankitsainiankit",
          "author_url": "",
          "post_date": "02/20/2020 17:27:23",
          "content": "<p>As far as I know, they mean you can't install them using the standard method of installing custom packages in Kaggle notebook. You'll have to manually load them through the dataset.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 751940,
      "author_name": "humananalog",
      "author_url": "",
      "post_date": "02/20/2020 16:54:48",
      "content": "<p>This is why I use BlazeFace, it's not perfect but it's fast. <a href=\"https://www.kaggle.com/humananalog/starter-blazeface-pytorch\">https://www.kaggle.com/humananalog/starter-blazeface-pytorch</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 751969,
          "author_name": "ankitsainiankit",
          "author_url": "",
          "post_date": "02/20/2020 17:26:11",
          "content": "<p>Thank you for your suggestion. I'll definitely look into it.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "746696": "How are you dealing with the time constraints of submission?  and what's your priority accuracy of faces or more frames for prediction\nIn my case, It takes almost 16s to detect face in all the 300 frames of a video(including video read time). Which is not good enough for submission. Currently, I've limited myself to 8-24 frames for each video in submission.\nBTW I'm using OpenCV face detection with cnn model.",
    "751934": "Are you even allowed to use openCV? \n\nIn the code requirements is stated: No custom packages enabled in your submission notebook",
    "751940": "This is why I use BlazeFace, it's not perfect but it's fast. https://www.kaggle.com/humananalog/starter-blazeface-pytorch",
    "751969": "Thank you for your suggestion. I'll definitely look into it.",
    "751970": "As far as I know, they mean you can't install them using the standard method of installing custom packages in Kaggle notebook. You'll have to manually load them through the dataset."
  },
  "source": "meta"
}