{
  "id": 506823,
  "title": "Third Place Solution",
  "url": "/competitions/ibiohash-2024-fgvc11/discussion/506823",
  "author_name": "",
  "post_date": "2024-05-23T11:07:53.980734800Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Thanks to the organizers for organizing this interesting contest to learn a lot in the field of hash retrieval.<br>\nSummarize<br>\nIn the following I will briefly describe our program.</p>\n<p>We use unicom to train the classification task and obtain a robust baseline.<br>\nThen we perform feature extraction on the query image and the gallery image, and use the faiss vector library for storage and retrieval.<br>\nFinally we generate a 48-bit hash code based on the reverse of the query results.</p>\n<p>Here is the link to our main code base in the GitHub code repository<br>\ngit@github.com:wangyijunlyy/iBioHash-fgvc-unicom.git</p>",
  "messages": [
    {
      "id": "2830794",
      "postDate": "05/23/2024 11:07:53",
      "content": "<p>Thanks to the organizers for organizing this interesting contest to learn a lot in the field of hash retrieval.<br>\nSummarize<br>\nIn the following I will briefly describe our program.</p>\n<p>We use unicom to train the classification task and obtain a robust baseline.<br>\nThen we perform feature extraction on the query image and the gallery image, and use the faiss vector library for storage and retrieval.<br>\nFinally we generate a 48-bit hash code based on the reverse of the query results.</p>\n<p>Here is the link to our main code base in the GitHub code repository<br>\ngit@github.com:wangyijunlyy/iBioHash-fgvc-unicom.git</p>",
      "rawMarkdown": "Thanks to the organizers for organizing this interesting contest to learn a lot in the field of hash retrieval.\nSummarize\nIn the following I will briefly describe our program.\n\nWe use unicom to train the classification task and obtain a robust baseline.\nThen we perform feature extraction on the query image and the gallery image, and use the faiss vector library for storage and retrieval.\nFinally we generate a 48-bit hash code based on the reverse of the query results.\n\nHere is the link to our main code base in the GitHub code repository\ngit@github.com:wangyijunlyy/iBioHash-fgvc-unicom.git",
      "votes": null
    },
    {
      "id": "2832936",
      "postDate": "05/24/2024 00:56:04",
      "content": "<p>Isn't Second place (instead of 3rd solution?) Check the LB wwwyyyjjj (It's a lot of triples : )<br>\nCongratulations and thanks for the GitHub link.</p>",
      "rawMarkdown": "Isn't Second place (instead of 3rd solution?) Check the LB wwwyyyjjj (It's a lot of triples : )\nCongratulations and thanks for the GitHub link.",
      "votes": null
    },
    {
      "id": "2833073",
      "postDate": "05/24/2024 04:19:41",
      "content": "<p>Final results are based on the 48-bits hash code. After checking, team 'Duong Anh Kiet' is better. Thank you for your support!</p>",
      "rawMarkdown": "Final results are based on the 48-bits hash code. After checking, team 'Duong Anh Kiet' is better. Thank you for your support!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2832936,
      "author_name": "mpwolke",
      "author_url": "",
      "post_date": "05/24/2024 00:56:04",
      "content": "<p>Isn't Second place (instead of 3rd solution?) Check the LB wwwyyyjjj (It's a lot of triples : )<br>\nCongratulations and thanks for the GitHub link.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2833073,
          "author_name": "aassxun",
          "author_url": "",
          "post_date": "05/24/2024 04:19:41",
          "content": "<p>Final results are based on the 48-bits hash code. After checking, team 'Duong Anh Kiet' is better. Thank you for your support!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2830794": "Thanks to the organizers for organizing this interesting contest to learn a lot in the field of hash retrieval.\nSummarize\nIn the following I will briefly describe our program.\n\nWe use unicom to train the classification task and obtain a robust baseline.\nThen we perform feature extraction on the query image and the gallery image, and use the faiss vector library for storage and retrieval.\nFinally we generate a 48-bit hash code based on the reverse of the query results.\n\nHere is the link to our main code base in the GitHub code repository\ngit@github.com:wangyijunlyy/iBioHash-fgvc-unicom.git",
    "2832936": "Isn't Second place (instead of 3rd solution?) Check the LB wwwyyyjjj (It's a lot of triples : )\nCongratulations and thanks for the GitHub link.",
    "2833073": "Final results are based on the 48-bits hash code. After checking, team 'Duong Anh Kiet' is better. Thank you for your support!"
  },
  "source": "meta"
}