{
  "id": 499101,
  "title": "[help] score 0.00 by just a few change in detect_keypoints",
  "url": "/competitions/image-matching-challenge-2024/discussion/499101",
  "author_name": "",
  "post_date": "2024-04-30T16:40:23.756788300Z",
  "votes": 5,
  "comment_count": 8,
  "views": 0,
  "content": "<p>I have a trouble in <strong>detect_keypoints</strong> function.<br>\nBy just adding a few lines of code to handle numpy array, I got 0.00 score(save&amp;run successfully).<br>\nIt run successfully and validation score of test/church is around 0.13 in both kaggle notebook and my local env.<br>\nThe submission.csv when save&amp;run is normal and almost same as other published notebook.<br>\nHowever, I got 0.00 when submitting. Are there any advice for this? The problem is around dealing with .h5 file?<br>\nSubmission time is too short,  ~10minutes while it took 10minutes when save&amp;run. Strage ? </p>\n<pre><code>\nkpts =  some_func(features[])\n kpts.shape == features[].shape\n (kpts) == (features[])\nf_keypoints[key] = kpts\n</code></pre>",
  "messages": [
    {
      "id": "2785155",
      "postDate": "04/30/2024 16:40:23",
      "content": "<p>I have a trouble in <strong>detect_keypoints</strong> function.<br>\nBy just adding a few lines of code to handle numpy array, I got 0.00 score(save&amp;run successfully).<br>\nIt run successfully and validation score of test/church is around 0.13 in both kaggle notebook and my local env.<br>\nThe submission.csv when save&amp;run is normal and almost same as other published notebook.<br>\nHowever, I got 0.00 when submitting. Are there any advice for this? The problem is around dealing with .h5 file?<br>\nSubmission time is too short,  ~10minutes while it took 10minutes when save&amp;run. Strage ? </p>\n<pre><code>\nkpts =  some_func(features[])\n kpts.shape == features[].shape\n (kpts) == (features[])\nf_keypoints[key] = kpts\n</code></pre>",
      "rawMarkdown": "I have a trouble in **detect_keypoints** function.\nBy just adding a few lines of code to handle numpy array, I got 0.00 score(save&run successfully).\nIt run successfully and validation score of test/church is around 0.13 in both kaggle notebook and my local env.\nThe submission.csv when save&run is normal and almost same as other published notebook.\nHowever, I got 0.00 when submitting. Are there any advice for this? The problem is around dealing with .h5 file?\nSubmission time is too short,  ~10minutes while it took 10minutes when save&run. Strage ? \n\n```python\n# in the detect_keypoints function\nkpts =  some_func(features[\"keypoints\"])\nassert kpts.shape == features[\"keypoints\"].shape\nassert type(kpts) == type(features[\"keypoints\"])\nf_keypoints[key] = kpts\n\n```",
      "votes": null
    },
    {
      "id": "2785526",
      "postDate": "04/30/2024 21:11:12",
      "content": "<p>i too faced it, it probably indicates that algo isn't detecting enough keypoints</p>",
      "rawMarkdown": "i too faced it, it probably indicates that algo isn't detecting enough keypoints",
      "votes": null
    },
    {
      "id": "2785612",
      "postDate": "04/30/2024 23:42:29",
      "content": "<p>Thanks for commenting.</p>\n<p>Did your algo work well in train data ?<br>\nIn my case, the validation score with all train scene data sampled up to 110 is higher than published baseline.</p>",
      "rawMarkdown": "Thanks for commenting.\n\nDid your algo work well in train data ?\nIn my case, the validation score with all train scene data sampled up to 110 is higher than published baseline.",
      "votes": null
    },
    {
      "id": "2785701",
      "postDate": "05/01/2024 02:01:22",
      "content": "<p>What did you sample up to 110? I didn't get the question tbh</p>",
      "rawMarkdown": "What did you sample up to 110? I didn't get the question tbh",
      "votes": null
    },
    {
      "id": "2785730",
      "postDate": "05/01/2024 02:25:57",
      "content": "<p>In my local valuation, a scene of train data is sampled up to N=110 as others do.</p>",
      "rawMarkdown": "In my local valuation, a scene of train data is sampled up to N=110 as others do.",
      "votes": null
    },
    {
      "id": "2785955",
      "postDate": "05/01/2024 05:10:07",
      "content": "<p>What is N here😶</p>",
      "rawMarkdown": "What is N here😶",
      "votes": null
    },
    {
      "id": "2800792",
      "postDate": "05/08/2024 11:07:32",
      "content": "<p>I have meet the same question.Did you solve it?Thanks for you answer</p>",
      "rawMarkdown": "I have meet the same question.Did you solve it?Thanks for you answer",
      "votes": null
    },
    {
      "id": "2800841",
      "postDate": "05/08/2024 11:32:52",
      "content": "<p>Yes. In my case, I found that keypoints were not detected in submission time by heavy trials and errors.</p>",
      "rawMarkdown": "Yes. In my case, I found that keypoints were not detected in submission time by heavy trials and errors.",
      "votes": null
    },
    {
      "id": "2800887",
      "postDate": "05/08/2024 11:51:40",
      "content": "<p>Pardon me<br>\nHow can you face the same issue<br>\nYou are ranked 11th😂</p>",
      "rawMarkdown": "Pardon me\nHow can you face the same issue\nYou are ranked 11th😂",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2785526,
      "author_name": "bhavesjain",
      "author_url": "",
      "post_date": "04/30/2024 21:11:12",
      "content": "<p>i too faced it, it probably indicates that algo isn't detecting enough keypoints</p>",
      "votes": null,
      "replies": [
        {
          "id": 2785612,
          "author_name": "clearwaterkzk",
          "author_url": "",
          "post_date": "04/30/2024 23:42:29",
          "content": "<p>Thanks for commenting.</p>\n<p>Did your algo work well in train data ?<br>\nIn my case, the validation score with all train scene data sampled up to 110 is higher than published baseline.</p>",
          "votes": null,
          "replies": [
            {
              "id": 2785701,
              "author_name": "bhavesjain",
              "author_url": "",
              "post_date": "05/01/2024 02:01:22",
              "content": "<p>What did you sample up to 110? I didn't get the question tbh</p>",
              "votes": null,
              "replies": [
                {
                  "id": 2785730,
                  "author_name": "clearwaterkzk",
                  "author_url": "",
                  "post_date": "05/01/2024 02:25:57",
                  "content": "<p>In my local valuation, a scene of train data is sampled up to N=110 as others do.</p>",
                  "votes": null,
                  "replies": [
                    {
                      "id": 2785955,
                      "author_name": "bhavesjain",
                      "author_url": "",
                      "post_date": "05/01/2024 05:10:07",
                      "content": "<p>What is N here😶</p>",
                      "votes": null,
                      "replies": []
                    }
                  ]
                }
              ]
            }
          ]
        }
      ]
    },
    {
      "id": 2800792,
      "author_name": "ustcfatcat",
      "author_url": "",
      "post_date": "05/08/2024 11:07:32",
      "content": "<p>I have meet the same question.Did you solve it?Thanks for you answer</p>",
      "votes": null,
      "replies": [
        {
          "id": 2800841,
          "author_name": "clearwaterkzk",
          "author_url": "",
          "post_date": "05/08/2024 11:32:52",
          "content": "<p>Yes. In my case, I found that keypoints were not detected in submission time by heavy trials and errors.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 2800887,
          "author_name": "bhavesjain",
          "author_url": "",
          "post_date": "05/08/2024 11:51:40",
          "content": "<p>Pardon me<br>\nHow can you face the same issue<br>\nYou are ranked 11th😂</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2785155": "I have a trouble in **detect_keypoints** function.\nBy just adding a few lines of code to handle numpy array, I got 0.00 score(save&run successfully).\nIt run successfully and validation score of test/church is around 0.13 in both kaggle notebook and my local env.\nThe submission.csv when save&run is normal and almost same as other published notebook.\nHowever, I got 0.00 when submitting. Are there any advice for this? The problem is around dealing with .h5 file?\nSubmission time is too short,  ~10minutes while it took 10minutes when save&run. Strage ? \n\n```python\n# in the detect_keypoints function\nkpts =  some_func(features[\"keypoints\"])\nassert kpts.shape == features[\"keypoints\"].shape\nassert type(kpts) == type(features[\"keypoints\"])\nf_keypoints[key] = kpts\n\n```",
    "2785526": "i too faced it, it probably indicates that algo isn't detecting enough keypoints",
    "2785612": "Thanks for commenting.\n\nDid your algo work well in train data ?\nIn my case, the validation score with all train scene data sampled up to 110 is higher than published baseline.",
    "2785701": "What did you sample up to 110? I didn't get the question tbh",
    "2785730": "In my local valuation, a scene of train data is sampled up to N=110 as others do.",
    "2785955": "What is N here😶",
    "2800792": "I have meet the same question.Did you solve it?Thanks for you answer",
    "2800841": "Yes. In my case, I found that keypoints were not detected in submission time by heavy trials and errors.",
    "2800887": "Pardon me\nHow can you face the same issue\nYou are ranked 11th😂"
  },
  "source": "meta"
}