{
  "id": 77258,
  "title": "SLF vs. CNN features",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/77258",
  "author_name": "",
  "post_date": "2019-01-11T01:02:28.318508800Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I'm wondering if anyone has made a submission based on SLF classification results. (Loc-CAT inference or other models) I might be wrong, but I remember from the paper that they claim to have achieved 0.72 F1.</p>\n\n<p><a href=\"https://www.ncbi.nlm.nih.gov/pubmed/30125267\">https://www.ncbi.nlm.nih.gov/pubmed/30125267</a> </p>",
  "messages": [
    {
      "id": "453940",
      "postDate": "01/11/2019 01:02:28",
      "content": "<p>I'm wondering if anyone has made a submission based on SLF classification results. (Loc-CAT inference or other models) I might be wrong, but I remember from the paper that they claim to have achieved 0.72 F1.</p>\n\n<p><a href=\"https://www.ncbi.nlm.nih.gov/pubmed/30125267\">https://www.ncbi.nlm.nih.gov/pubmed/30125267</a> </p>",
      "rawMarkdown": "I'm wondering if anyone has made a submission based on SLF classification results. (Loc-CAT inference or other models) I might be wrong, but I remember from the paper that they claim to have achieved 0.72 F1.\n\nhttps://www.ncbi.nlm.nih.gov/pubmed/30125267",
      "votes": null
    },
    {
      "id": "454234",
      "postDate": "01/11/2019 09:27:21",
      "content": "<p>I tried, Loc-CAT was nightmarish to get working and I gave up.</p>\n\n<p>I also tried using the inpainting method from Lu et al. (2018) \n<a href=\"https://github.com/alexxijielu/paired_cell_inpainting#human-model\">https://github.com/alexxijielu/paired_cell_inpainting#human-model</a></p>\n\n<p>That seemed promising, but it needed a bit more retraining than I had skill for. I did end up with ~4 million single cell crops from the training set + hpa18, dunno what I'm going to do with them all...</p>",
      "rawMarkdown": "I tried, Loc-CAT was nightmarish to get working and I gave up.\n\nI also tried using the inpainting method from Lu et al. (2018) \nhttps://github.com/alexxijielu/paired_cell_inpainting#human-model\n\nThat seemed promising, but it needed a bit more retraining than I had skill for. I did end up with ~4 million single cell crops from the training set + hpa18, dunno what I'm going to do with them all...",
      "votes": null
    },
    {
      "id": "454631",
      "postDate": "01/11/2019 22:44:34",
      "content": "<p>So were you able to generate a submission from Loc-CAT? I only did a few on training set, some are correct, some are wrong.</p>",
      "rawMarkdown": "So were you able to generate a submission from Loc-CAT? I only did a few on training set, some are correct, some are wrong.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 454234,
      "author_name": "interneuron",
      "author_url": "",
      "post_date": "01/11/2019 09:27:21",
      "content": "<p>I tried, Loc-CAT was nightmarish to get working and I gave up.</p>\n\n<p>I also tried using the inpainting method from Lu et al. (2018) \n<a href=\"https://github.com/alexxijielu/paired_cell_inpainting#human-model\">https://github.com/alexxijielu/paired_cell_inpainting#human-model</a></p>\n\n<p>That seemed promising, but it needed a bit more retraining than I had skill for. I did end up with ~4 million single cell crops from the training set + hpa18, dunno what I'm going to do with them all...</p>",
      "votes": null,
      "replies": [
        {
          "id": 454631,
          "author_name": "alexanderliao",
          "author_url": "",
          "post_date": "01/11/2019 22:44:34",
          "content": "<p>So were you able to generate a submission from Loc-CAT? I only did a few on training set, some are correct, some are wrong.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "453940": "I'm wondering if anyone has made a submission based on SLF classification results. (Loc-CAT inference or other models) I might be wrong, but I remember from the paper that they claim to have achieved 0.72 F1.\n\nhttps://www.ncbi.nlm.nih.gov/pubmed/30125267",
    "454234": "I tried, Loc-CAT was nightmarish to get working and I gave up.\n\nI also tried using the inpainting method from Lu et al. (2018) \nhttps://github.com/alexxijielu/paired_cell_inpainting#human-model\n\nThat seemed promising, but it needed a bit more retraining than I had skill for. I did end up with ~4 million single cell crops from the training set + hpa18, dunno what I'm going to do with them all...",
    "454631": "So were you able to generate a submission from Loc-CAT? I only did a few on training set, some are correct, some are wrong."
  },
  "source": "meta"
}