{
  "id": 585225,
  "title": "Getting started",
  "url": "/competitions/aibt-hackathon-2025-image-retrieval-public/discussion/585225",
  "author_name": "Yannick Prudent",
  "post_date": "2025-06-18T22:37:07.478000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Not sure where to begin? You can try deep metric learning with triplet loss:</p>\n<ul>\n<li><p><strong>Triplet Network Paper</strong><br>\n<a href=\"https://arxiv.org/abs/1412.6622\" target=\"_blank\"> “Deep Metric Learning Using Triplet Network”</a> by Hoffer &amp; Ailon (ICLR 2015) introduces the triplet-based embedding framework for learning distances between samples .</p></li>\n<li><p><strong>Data Loader Template</strong><br>\nA concise PyTorch implementation for triplet sampling (credits to our colleague Adil Zouitine), ready to integrate into your training loop: <a href=\"https://gist.github.com/AdilZouitine/0545520c8f377c1adc91a0138758c22b\" target=\"_blank\">https://gist.github.com/AdilZouitine/0545520c8f377c1adc91a0138758c22b</a>.</p></li>\n<li><p><strong>Related Competition</strong><br>\nCheck out the discussion from the Humpback Whale Identification challenge on Kaggle, where many teams successfully applied triplet-loss approaches: <a href=\"https://www.kaggle.com/c/humpback-whale-identification/discussion?sort=recent-comments\" target=\"_blank\">https://www.kaggle.com/c/humpback-whale-identification/discussion?sort=recent-comments</a></p></li>\n</ul>\n<p>Good luck!</p>",
  "messages": [
    {
      "id": 3227362,
      "postDate": "2025-06-18T22:37:07.480Z",
      "content": "<p>Not sure where to begin? You can try deep metric learning with triplet loss:</p>\n<ul>\n<li><p><strong>Triplet Network Paper</strong><br>\n<a href=\"https://arxiv.org/abs/1412.6622\" target=\"_blank\"> “Deep Metric Learning Using Triplet Network”</a> by Hoffer &amp; Ailon (ICLR 2015) introduces the triplet-based embedding framework for learning distances between samples .</p></li>\n<li><p><strong>Data Loader Template</strong><br>\nA concise PyTorch implementation for triplet sampling (credits to our colleague Adil Zouitine), ready to integrate into your training loop: <a href=\"https://gist.github.com/AdilZouitine/0545520c8f377c1adc91a0138758c22b\" target=\"_blank\">https://gist.github.com/AdilZouitine/0545520c8f377c1adc91a0138758c22b</a>.</p></li>\n<li><p><strong>Related Competition</strong><br>\nCheck out the discussion from the Humpback Whale Identification challenge on Kaggle, where many teams successfully applied triplet-loss approaches: <a href=\"https://www.kaggle.com/c/humpback-whale-identification/discussion?sort=recent-comments\" target=\"_blank\">https://www.kaggle.com/c/humpback-whale-identification/discussion?sort=recent-comments</a></p></li>\n</ul>\n<p>Good luck!</p>",
      "rawMarkdown": "Not sure where to begin? You can try deep metric learning with triplet loss:\n\n* **Triplet Network Paper**\n [ “Deep Metric Learning Using Triplet Network”](https://arxiv.org/abs/1412.6622) by Hoffer & Ailon (ICLR 2015) introduces the triplet-based embedding framework for learning distances between samples .\n\n* **Data Loader Template**\n  A concise PyTorch implementation for triplet sampling (credits to our colleague Adil Zouitine), ready to integrate into your training loop: [https://gist.github.com/AdilZouitine/0545520c8f377c1adc91a0138758c22b](https://gist.github.com/AdilZouitine/0545520c8f377c1adc91a0138758c22b).\n\n* **Related Competition**\n  Check out the discussion from the Humpback Whale Identification challenge on Kaggle, where many teams successfully applied triplet-loss approaches: [https://www.kaggle.com/c/humpback-whale-identification/discussion?sort=recent-comments](https://www.kaggle.com/c/humpback-whale-identification/discussion?sort=recent-comments)\n\nGood luck!"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3227362": "Not sure where to begin? You can try deep metric learning with triplet loss:\n\n* **Triplet Network Paper**\n [ “Deep Metric Learning Using Triplet Network”](https://arxiv.org/abs/1412.6622) by Hoffer & Ailon (ICLR 2015) introduces the triplet-based embedding framework for learning distances between samples .\n\n* **Data Loader Template**\n  A concise PyTorch implementation for triplet sampling (credits to our colleague Adil Zouitine), ready to integrate into your training loop: [https://gist.github.com/AdilZouitine/0545520c8f377c1adc91a0138758c22b](https://gist.github.com/AdilZouitine/0545520c8f377c1adc91a0138758c22b).\n\n* **Related Competition**\n  Check out the discussion from the Humpback Whale Identification challenge on Kaggle, where many teams successfully applied triplet-loss approaches: [https://www.kaggle.com/c/humpback-whale-identification/discussion?sort=recent-comments](https://www.kaggle.com/c/humpback-whale-identification/discussion?sort=recent-comments)\n\nGood luck!"
  }
}