{
  "id": 117998,
  "title": "Starting from Basic ",
  "url": "/competitions/pku-autonomous-driving/discussion/117998",
  "author_name": "Nirjhar Roy",
  "post_date": "2019-11-19T05:51:55.630000",
  "votes": 13,
  "comment_count": 14,
  "views": 0,
  "content": "<p>Hi All , </p>\n\n<p>After the cloud comp , I want to now try to learn in this competition .As I said few days back , that it feels like complex problem to solve and visualize . I dont know anything about car or driving so conceptually that is added difficulty . I am trying to grasp the basic concepts. Just now I took a picture of the novel I am reading . The edge of the book facing me is little down than the farther edge , if i consider floor as a horizontal plain.Is that pitch ? \nThe edge of the book that is at my right is little down than the left side , is that roll ? \nIf i see the book through my mobile camera , it also looks like the edge of the book that is farther from me is shifted to the right a little bit more in the horizontal plain . Is that Yaw ? <br>\nIs the X,Y,Z parameter relative to the camera ? as shown in the picture ? \nIf this is true , can I assume that X,Y,Z is dependent on camera and yaw,pitch , roll is the property of individual cars not dependent on relative distance from camera ? </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F7531a75893e4787c32b1c9721ee827e0%2FVerify_PKU.png?generation=1574142702564809&amp;alt=media\" alt=\"\"></p>\n\n<p>What is the use of pitch and roll in real life car driving ? Do you see if the car is rolling or i guess when the car in front put sudden break , you guess that the pitch of the car might come into play , but are those important parameter for driving ?</p>",
  "messages": [
    {
      "id": 676370,
      "postDate": "2019-11-19T05:51:55.630Z",
      "content": "<p>Hi All , </p>\n\n<p>After the cloud comp , I want to now try to learn in this competition .As I said few days back , that it feels like complex problem to solve and visualize . I dont know anything about car or driving so conceptually that is added difficulty . I am trying to grasp the basic concepts. Just now I took a picture of the novel I am reading . The edge of the book facing me is little down than the farther edge , if i consider floor as a horizontal plain.Is that pitch ? \nThe edge of the book that is at my right is little down than the left side , is that roll ? \nIf i see the book through my mobile camera , it also looks like the edge of the book that is farther from me is shifted to the right a little bit more in the horizontal plain . Is that Yaw ? <br>\nIs the X,Y,Z parameter relative to the camera ? as shown in the picture ? \nIf this is true , can I assume that X,Y,Z is dependent on camera and yaw,pitch , roll is the property of individual cars not dependent on relative distance from camera ? </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F7531a75893e4787c32b1c9721ee827e0%2FVerify_PKU.png?generation=1574142702564809&amp;alt=media\" alt=\"\"></p>\n\n<p>What is the use of pitch and roll in real life car driving ? Do you see if the car is rolling or i guess when the car in front put sudden break , you guess that the pitch of the car might come into play , but are those important parameter for driving ?</p>",
      "rawMarkdown": "Hi All , \n\nAfter the cloud comp , I want to now try to learn in this competition .As I said few days back , that it feels like complex problem to solve and visualize . I dont know anything about car or driving so conceptually that is added difficulty . I am trying to grasp the basic concepts. Just now I took a picture of the novel I am reading . The edge of the book facing me is little down than the farther edge , if i consider floor as a horizontal plain.Is that pitch ? \nThe edge of the book that is at my right is little down than the left side , is that roll ? \nIf i see the book through my mobile camera , it also looks like the edge of the book that is farther from me is shifted to the right a little bit more in the horizontal plain . Is that Yaw ?  \nIs the X,Y,Z parameter relative to the camera ? as shown in the picture ? \nIf this is true , can I assume that X,Y,Z is dependent on camera and yaw,pitch , roll is the property of individual cars not dependent on relative distance from camera ? \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F7531a75893e4787c32b1c9721ee827e0%2FVerify_PKU.png?generation=1574142702564809&amp;alt=media)\n\nWhat is the use of pitch and roll in real life car driving ? Do you see if the car is rolling or i guess when the car in front put sudden break , you guess that the pitch of the car might come into play , but are those important parameter for driving ?",
      "votes": 13
    },
    {
      "id": 676738,
      "postDate": "2019-11-19T13:24:35.810Z",
      "content": "<p>Here is my comprehension of the car angles. They are 100% based on what the competition is looking for and doesn't correspond to what the actual definitions are. (They are also in the order that they should be submitted in)\n- Yaw is the lean of the car. For example going up or down a hill (when directly in front of camera) or braking as in your example. Yaw should be 0 most of the time but the camera is setup in a way where the angle is offset by about .15.\n- Pitch is the value for turn angle: -3.1 (sometimes 3.1) for facing camera, 0 for facing away, 1.6 for facing left, -1.6 for facing right.\n- Roll is the sideways lean of the car. If the car got into an accident and rolled over then roll would be 0. Roll is almost always -3.1 but can sometimes be 3.1. (Refer to this: <a href=\"https://www.kaggle.com/c/pku-autonomous-driving/discussion/116423#latest-672019\">https://www.kaggle.com/c/pku-autonomous-driving/discussion/116423#latest-672019</a> )</p>",
      "rawMarkdown": "Here is my comprehension of the car angles. They are 100% based on what the competition is looking for and doesn't correspond to what the actual definitions are. (They are also in the order that they should be submitted in)\n- Yaw is the lean of the car. For example going up or down a hill (when directly in front of camera) or braking as in your example. Yaw should be 0 most of the time but the camera is setup in a way where the angle is offset by about .15.\n- Pitch is the value for turn angle: -3.1 (sometimes 3.1) for facing camera, 0 for facing away, 1.6 for facing left, -1.6 for facing right.\n- Roll is the sideways lean of the car. If the car got into an accident and rolled over then roll would be 0. Roll is almost always -3.1 but can sometimes be 3.1. (Refer to this: https://www.kaggle.com/c/pku-autonomous-driving/discussion/116423#latest-672019 )",
      "votes": 9,
      "replies": [
        {
          "id": 676754,
          "postDate": "2019-11-19T13:34:53.647Z",
          "content": "<p>Now I am clear , what people were discussing about  in forum :  that yaw and pitch got exchanged in the competition data :) . I could not comprehend it so long .  Thanks a lot for clarifying . <br>\nBut as a driver I think yaw ( in this competition pitch) should be more important right ? Then comes pitch . I am still trying to understand what would roll be useful for , in case of a human driver . </p>",
          "rawMarkdown": "Now I am clear , what people were discussing about  in forum :  that yaw and pitch got exchanged in the competition data :) . I could not comprehend it so long .  Thanks a lot for clarifying .  \nBut as a driver I think yaw ( in this competition pitch) should be more important right ? Then comes pitch . I am still trying to understand what would roll be useful for , in case of a human driver . "
        },
        {
          "id": 676766,
          "postDate": "2019-11-19T13:42:11.773Z",
          "content": "<p>I agree, really only the turning angle is important. But in this competition the makers want to get every possible bit of information out of the images (even if it's unhelpful in 99.9% of circumstances)</p>",
          "rawMarkdown": "I agree, really only the turning angle is important. But in this competition the makers want to get every possible bit of information out of the images (even if it's unhelpful in 99.9% of circumstances)"
        },
        {
          "id": 697190,
          "postDate": "2019-12-17T15:54:10.880Z",
          "content": "<p>Thanks <a href=\"/greatgamedota\">@greatgamedota</a> , this is by far the clearest explanation I have seen.  Really appreciate your diligence to help the community!</p>",
          "rawMarkdown": "Thanks @greatgamedota , this is by far the clearest explanation I have seen.  Really appreciate your diligence to help the community!"
        },
        {
          "id": 697883,
          "postDate": "2019-12-18T14:09:59.913Z",
          "content": "<p><a href=\"/greatgamedota\">@greatgamedota</a> <br>\n1) please take some more effort to help understand this.\n<a href=\"https://www.kaggle.com/seshadrikolluri/vehicle-angle-prediction-understanding-eda\">https://www.kaggle.com/seshadrikolluri/vehicle-angle-prediction-understanding-eda</a>\nif you look at the starting pictures of kernel . your pitch and yaw are seem to have inverted meanings. Its Yaw which is side wise turn ( in case of road turn) meaning rotation across vertical axis and pitch is hz axis which is useful when car is driving on up or down plane  wrt to camera ,and roll is if suppose there is a  straight road such that one side is above or below the other side of road . Please validate my understanding</p>\n\n<p>2) if suppose road is very straight  ,in that case will all these angles be zero for a car which is at left  camera and ahead of the camera.</p>",
          "rawMarkdown": "@greatgamedota  \n1) please take some more effort to help understand this.\nhttps://www.kaggle.com/seshadrikolluri/vehicle-angle-prediction-understanding-eda\nif you look at the starting pictures of kernel . your pitch and yaw are seem to have inverted meanings. Its Yaw which is side wise turn ( in case of road turn) meaning rotation across vertical axis and pitch is hz axis which is useful when car is driving on up or down plane  wrt to camera ,and roll is if suppose there is a  straight road such that one side is above or below the other side of road . Please validate my understanding\n\n2) if suppose road is very straight  ,in that case will all these angles be zero for a car which is at left  camera and ahead of the camera.\n"
        },
        {
          "id": 697934,
          "postDate": "2019-12-18T15:09:09.553Z",
          "content": "<ol>\n<li>You're correct, it is just the competition organizers mislabeled yaw and pitch so they mean opposite things to what is generally agreed they do.</li>\n<li>Everything is explained in the original post. Where the car is positioned doesn't have anything to do with the pose angles.</li>\n</ol>",
          "rawMarkdown": "1. You're correct, it is just the competition organizers mislabeled yaw and pitch so they mean opposite things to what is generally agreed they do.\n2. Everything is explained in the original post. Where the car is positioned doesn't have anything to do with the pose angles."
        },
        {
          "id": 705834,
          "postDate": "2019-12-29T15:01:47.733Z",
          "content": "<p><a href=\"/greatgamedota\">@greatgamedota</a> \nwhile building the label we shouldnt be bothered about interchanging Y and P \nIs competition evaulating Y as P  and P as Y during submission ?\nsorry for my confusion..\nstill trying to understand why is it still ok to train with same sequence of labels..</p>",
          "rawMarkdown": "@greatgamedota \nwhile building the label we shouldnt be bothered about interchanging Y and P \nIs competition evaulating Y as P  and P as Y during submission ?\nsorry for my confusion..\nstill trying to understand why is it still ok to train with same sequence of labels.."
        },
        {
          "id": 705862,
          "postDate": "2019-12-29T15:37:58.397Z",
          "content": "<p>Yaw and Pitch are very different so interchanging them will lower your score drastically.</p>",
          "rawMarkdown": "Yaw and Pitch are very different so interchanging them will lower your score drastically."
        },
        {
          "id": 705874,
          "postDate": "2019-12-29T15:51:33.520Z",
          "content": "<p>are the angles of Y/p/R given are in degrees or radians..\nsorry it may be mentioned some where m unable to find that out.</p>",
          "rawMarkdown": "are the angles of Y/p/R given are in degrees or radians..\nsorry it may be mentioned some where m unable to find that out."
        },
        {
          "id": 705886,
          "postDate": "2019-12-29T16:20:36.223Z",
          "content": "<p>Radians: <a href=\"https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation\">https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation</a></p>",
          "rawMarkdown": "Radians: https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation"
        }
      ]
    },
    {
      "id": 676672,
      "postDate": "2019-11-19T12:09:27.673Z",
      "content": "<p>i am new man too, you are right.  </p>",
      "rawMarkdown": "i am new man too, you are right.  "
    },
    {
      "id": 676978,
      "postDate": "2019-11-19T17:12:17.517Z",
      "rawMarkdown": "",
      "isDeleted": true,
      "replies": [
        {
          "id": 676979,
          "postDate": "2019-11-19T17:16:34.987Z",
          "content": "<p>I'm not, the competition organizers did, what I wrote is what is correct for this comp</p>",
          "rawMarkdown": "I'm not, the competition organizers did, what I wrote is what is correct for this comp",
          "votes": 2
        },
        {
          "id": 676983,
          "postDate": "2019-11-19T17:25:04.647Z",
          "content": "<p>yeah, my bad, I read your message inattentively, sorry</p>",
          "rawMarkdown": "yeah, my bad, I read your message inattentively, sorry"
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 676738,
      "author_name": "GreatGameDota",
      "author_url": "",
      "post_date": "2019-11-19T13:24:35.810000",
      "content": "<p>Here is my comprehension of the car angles. They are 100% based on what the competition is looking for and doesn't correspond to what the actual definitions are. (They are also in the order that they should be submitted in)\n- Yaw is the lean of the car. For example going up or down a hill (when directly in front of camera) or braking as in your example. Yaw should be 0 most of the time but the camera is setup in a way where the angle is offset by about .15.\n- Pitch is the value for turn angle: -3.1 (sometimes 3.1) for facing camera, 0 for facing away, 1.6 for facing left, -1.6 for facing right.\n- Roll is the sideways lean of the car. If the car got into an accident and rolled over then roll would be 0. Roll is almost always -3.1 but can sometimes be 3.1. (Refer to this: <a href=\"https://www.kaggle.com/c/pku-autonomous-driving/discussion/116423#latest-672019\">https://www.kaggle.com/c/pku-autonomous-driving/discussion/116423#latest-672019</a> )</p>",
      "votes": 9,
      "replies": [
        {
          "id": 676754,
          "author_name": "Nirjhar Roy",
          "author_url": "",
          "post_date": "2019-11-19T13:34:53.647000",
          "content": "<p>Now I am clear , what people were discussing about  in forum :  that yaw and pitch got exchanged in the competition data :) . I could not comprehend it so long .  Thanks a lot for clarifying . <br>\nBut as a driver I think yaw ( in this competition pitch) should be more important right ? Then comes pitch . I am still trying to understand what would roll be useful for , in case of a human driver . </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 676766,
          "author_name": "GreatGameDota",
          "author_url": "",
          "post_date": "2019-11-19T13:42:11.773000",
          "content": "<p>I agree, really only the turning angle is important. But in this competition the makers want to get every possible bit of information out of the images (even if it's unhelpful in 99.9% of circumstances)</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 697190,
          "author_name": "dustinvs",
          "author_url": "",
          "post_date": "2019-12-17T15:54:10.880000",
          "content": "<p>Thanks <a href=\"/greatgamedota\">@greatgamedota</a> , this is by far the clearest explanation I have seen.  Really appreciate your diligence to help the community!</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 697883,
          "author_name": "Jaideep",
          "author_url": "",
          "post_date": "2019-12-18T14:09:59.913000",
          "content": "<p><a href=\"/greatgamedota\">@greatgamedota</a> <br>\n1) please take some more effort to help understand this.\n<a href=\"https://www.kaggle.com/seshadrikolluri/vehicle-angle-prediction-understanding-eda\">https://www.kaggle.com/seshadrikolluri/vehicle-angle-prediction-understanding-eda</a>\nif you look at the starting pictures of kernel . your pitch and yaw are seem to have inverted meanings. Its Yaw which is side wise turn ( in case of road turn) meaning rotation across vertical axis and pitch is hz axis which is useful when car is driving on up or down plane  wrt to camera ,and roll is if suppose there is a  straight road such that one side is above or below the other side of road . Please validate my understanding</p>\n\n<p>2) if suppose road is very straight  ,in that case will all these angles be zero for a car which is at left  camera and ahead of the camera.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 697934,
          "author_name": "GreatGameDota",
          "author_url": "",
          "post_date": "2019-12-18T15:09:09.553000",
          "content": "<ol>\n<li>You're correct, it is just the competition organizers mislabeled yaw and pitch so they mean opposite things to what is generally agreed they do.</li>\n<li>Everything is explained in the original post. Where the car is positioned doesn't have anything to do with the pose angles.</li>\n</ol>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 705834,
          "author_name": "Jaideep",
          "author_url": "",
          "post_date": "2019-12-29T15:01:47.733000",
          "content": "<p><a href=\"/greatgamedota\">@greatgamedota</a> \nwhile building the label we shouldnt be bothered about interchanging Y and P \nIs competition evaulating Y as P  and P as Y during submission ?\nsorry for my confusion..\nstill trying to understand why is it still ok to train with same sequence of labels..</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 705862,
          "author_name": "GreatGameDota",
          "author_url": "",
          "post_date": "2019-12-29T15:37:58.397000",
          "content": "<p>Yaw and Pitch are very different so interchanging them will lower your score drastically.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 705874,
          "author_name": "Jaideep",
          "author_url": "",
          "post_date": "2019-12-29T15:51:33.520000",
          "content": "<p>are the angles of Y/p/R given are in degrees or radians..\nsorry it may be mentioned some where m unable to find that out.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 705886,
          "author_name": "GreatGameDota",
          "author_url": "",
          "post_date": "2019-12-29T16:20:36.223000",
          "content": "<p>Radians: <a href=\"https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation\">https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation</a></p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 676672,
      "author_name": "GuoBaozi",
      "author_url": "",
      "post_date": "2019-11-19T12:09:27.673000",
      "content": "<p>i am new man too, you are right.  </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 676978,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-11-19T17:12:17.517000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 676979,
          "author_name": "GreatGameDota",
          "author_url": "",
          "post_date": "2019-11-19T17:16:34.987000",
          "content": "<p>I'm not, the competition organizers did, what I wrote is what is correct for this comp</p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 676983,
          "author_name": "Larionov Alexander ",
          "author_url": "",
          "post_date": "2019-11-19T17:25:04.647000",
          "content": "<p>yeah, my bad, I read your message inattentively, sorry</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "676370": "Hi All , \n\nAfter the cloud comp , I want to now try to learn in this competition .As I said few days back , that it feels like complex problem to solve and visualize . I dont know anything about car or driving so conceptually that is added difficulty . I am trying to grasp the basic concepts. Just now I took a picture of the novel I am reading . The edge of the book facing me is little down than the farther edge , if i consider floor as a horizontal plain.Is that pitch ? \nThe edge of the book that is at my right is little down than the left side , is that roll ? \nIf i see the book through my mobile camera , it also looks like the edge of the book that is farther from me is shifted to the right a little bit more in the horizontal plain . Is that Yaw ?  \nIs the X,Y,Z parameter relative to the camera ? as shown in the picture ? \nIf this is true , can I assume that X,Y,Z is dependent on camera and yaw,pitch , roll is the property of individual cars not dependent on relative distance from camera ? \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F7531a75893e4787c32b1c9721ee827e0%2FVerify_PKU.png?generation=1574142702564809&amp;alt=media)\n\nWhat is the use of pitch and roll in real life car driving ? Do you see if the car is rolling or i guess when the car in front put sudden break , you guess that the pitch of the car might come into play , but are those important parameter for driving ?",
    "676738": "Here is my comprehension of the car angles. They are 100% based on what the competition is looking for and doesn't correspond to what the actual definitions are. (They are also in the order that they should be submitted in)\n- Yaw is the lean of the car. For example going up or down a hill (when directly in front of camera) or braking as in your example. Yaw should be 0 most of the time but the camera is setup in a way where the angle is offset by about .15.\n- Pitch is the value for turn angle: -3.1 (sometimes 3.1) for facing camera, 0 for facing away, 1.6 for facing left, -1.6 for facing right.\n- Roll is the sideways lean of the car. If the car got into an accident and rolled over then roll would be 0. Roll is almost always -3.1 but can sometimes be 3.1. (Refer to this: https://www.kaggle.com/c/pku-autonomous-driving/discussion/116423#latest-672019 )",
    "676672": "i am new man too, you are right.  ",
    "676978": ""
  }
}