{
  "id": 80125,
  "title": "classical sift matching",
  "url": "/competitions/humpback-whale-identification/discussion/80125",
  "author_name": "",
  "post_date": "2019-02-10T21:46:38.641673700Z",
  "votes": 4,
  "comment_count": 7,
  "views": 0,
  "content": "<p><a href=\"http://demo.ipol.im/demo/my_affine_sift/\">http://demo.ipol.im/demo/my_affine_sift/</a></p>\n\n<p>here are the results:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469283/11247/asift.png\" alt=\"enter image description here\"></p>\n\n<p>maybe i should try magic leap superpoint CNN keypoint matching too!\n(note that superpoint is trained on synthetic warp image. in theory, we can train a superpoint for whale image using synthetic warp image too)</p>",
  "messages": [
    {
      "id": "469283",
      "postDate": "02/10/2019 21:46:38",
      "content": "<p><a href=\"http://demo.ipol.im/demo/my_affine_sift/\">http://demo.ipol.im/demo/my_affine_sift/</a></p>\n\n<p>here are the results:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469283/11247/asift.png\" alt=\"enter image description here\"></p>\n\n<p>maybe i should try magic leap superpoint CNN keypoint matching too!\n(note that superpoint is trained on synthetic warp image. in theory, we can train a superpoint for whale image using synthetic warp image too)</p>",
      "rawMarkdown": "http://demo.ipol.im/demo/my_affine_sift/\n\nhere are the results:\n\n  ![enter image description here][1]\n\n\nmaybe i should try magic leap superpoint CNN keypoint matching too!\n(note that superpoint is trained on synthetic warp image. in theory, we can train a superpoint for whale image using synthetic warp image too)\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469283/11247/asift.png",
      "votes": null
    },
    {
      "id": "469285",
      "postDate": "02/10/2019 21:50:33",
      "content": "<p>related:  <a href=\"https://github.com/tensorflow/models/tree/master/research/delf\">https://github.com/tensorflow/models/tree/master/research/delf</a></p>",
      "rawMarkdown": "related:  https://github.com/tensorflow/models/tree/master/research/delf",
      "votes": null
    },
    {
      "id": "469287",
      "postDate": "02/10/2019 21:57:41",
      "content": "<p>...</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469287/11248/asift1.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "...\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469287/11248/asift1.png",
      "votes": null
    },
    {
      "id": "469288",
      "postDate": "02/10/2019 22:01:37",
      "content": "<p>maybe can be used for labeling for learning deep view transform?</p>\n\n<p>or input for CNN to verify if it is a match or not?\ne.g. image1, image2,  keypoint_match ---&gt; [CNN] ---&gt; match or not</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469288/11249/asift2.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "maybe can be used for labeling for learning deep view transform?\n\nor input for CNN to verify if it is a match or not?\ne.g. image1, image2,  keypoint_match ---&gt; [CNN] ---&gt; match or not\n\n\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469288/11249/asift2.png",
      "votes": null
    },
    {
      "id": "469291",
      "postDate": "02/10/2019 22:09:33",
      "content": "<p>...</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469291/11250/asift3..png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "...\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469291/11250/asift3..png",
      "votes": null
    },
    {
      "id": "469297",
      "postDate": "02/10/2019 22:21:16",
      "content": "<p><a href=\"https://github.com/MagicLeapResearch/SuperPointPretrainedNetwork\">https://github.com/MagicLeapResearch/SuperPointPretrainedNetwork</a></p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469297/11251/superpoint.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "https://github.com/MagicLeapResearch/SuperPointPretrainedNetwork\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469297/11251/superpoint.png",
      "votes": null
    },
    {
      "id": "469374",
      "postDate": "02/11/2019 04:24:05",
      "content": "<p>Local feature matching is practical method and I want to exploit it actually. The most troubling aspect of this approach is handing new_whale. Thresholding number of inliers seems to be not enough.</p>",
      "rawMarkdown": "Local feature matching is practical method and I want to exploit it actually. The most troubling aspect of this approach is handing new_whale. Thresholding number of inliers seems to be not enough.",
      "votes": null
    },
    {
      "id": "469454",
      "postDate": "02/11/2019 08:42:18",
      "content": "<p>ASIFT from demo is extremely slow. If you extraction from ASIFT and then matching yourself, it is much faster. </p>\n\n<p>Or just use MODS :)\n<a href=\"https://github.com/ducha-aiki/mods-light-zmq\">https://github.com/ducha-aiki/mods-light-zmq</a></p>",
      "rawMarkdown": "ASIFT from demo is extremely slow. If you extraction from ASIFT and then matching yourself, it is much faster. \n\nOr just use MODS :)\nhttps://github.com/ducha-aiki/mods-light-zmq",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 469285,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "02/10/2019 21:50:33",
      "content": "<p>related:  <a href=\"https://github.com/tensorflow/models/tree/master/research/delf\">https://github.com/tensorflow/models/tree/master/research/delf</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 469374,
          "author_name": "toshik",
          "author_url": "",
          "post_date": "02/11/2019 04:24:05",
          "content": "<p>Local feature matching is practical method and I want to exploit it actually. The most troubling aspect of this approach is handing new_whale. Thresholding number of inliers seems to be not enough.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 469287,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "02/10/2019 21:57:41",
      "content": "<p>...</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469287/11248/asift1.png\" alt=\"enter image description here\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 469288,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "02/10/2019 22:01:37",
      "content": "<p>maybe can be used for labeling for learning deep view transform?</p>\n\n<p>or input for CNN to verify if it is a match or not?\ne.g. image1, image2,  keypoint_match ---&gt; [CNN] ---&gt; match or not</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469288/11249/asift2.png\" alt=\"enter image description here\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 469291,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "02/10/2019 22:09:33",
      "content": "<p>...</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469291/11250/asift3..png\" alt=\"enter image description here\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 469297,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "02/10/2019 22:21:16",
      "content": "<p><a href=\"https://github.com/MagicLeapResearch/SuperPointPretrainedNetwork\">https://github.com/MagicLeapResearch/SuperPointPretrainedNetwork</a></p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/469297/11251/superpoint.png\" alt=\"enter image description here\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 469454,
      "author_name": "oldufo",
      "author_url": "",
      "post_date": "02/11/2019 08:42:18",
      "content": "<p>ASIFT from demo is extremely slow. If you extraction from ASIFT and then matching yourself, it is much faster. </p>\n\n<p>Or just use MODS :)\n<a href=\"https://github.com/ducha-aiki/mods-light-zmq\">https://github.com/ducha-aiki/mods-light-zmq</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "469283": "http://demo.ipol.im/demo/my_affine_sift/\n\nhere are the results:\n\n  ![enter image description here][1]\n\n\nmaybe i should try magic leap superpoint CNN keypoint matching too!\n(note that superpoint is trained on synthetic warp image. in theory, we can train a superpoint for whale image using synthetic warp image too)\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469283/11247/asift.png",
    "469285": "related:  https://github.com/tensorflow/models/tree/master/research/delf",
    "469287": "...\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469287/11248/asift1.png",
    "469288": "maybe can be used for labeling for learning deep view transform?\n\nor input for CNN to verify if it is a match or not?\ne.g. image1, image2,  keypoint_match ---&gt; [CNN] ---&gt; match or not\n\n\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469288/11249/asift2.png",
    "469291": "...\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469291/11250/asift3..png",
    "469297": "https://github.com/MagicLeapResearch/SuperPointPretrainedNetwork\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/469297/11251/superpoint.png",
    "469374": "Local feature matching is practical method and I want to exploit it actually. The most troubling aspect of this approach is handing new_whale. Thresholding number of inliers seems to be not enough.",
    "469454": "ASIFT from demo is extremely slow. If you extraction from ASIFT and then matching yourself, it is much faster. \n\nOr just use MODS :)\nhttps://github.com/ducha-aiki/mods-light-zmq"
  },
  "source": "meta"
}