{
  "id": 327007,
  "title": "Just three steps to a bronze medal",
  "url": "/competitions/birdclef-2022/discussion/327007",
  "author_name": "Liuxwww",
  "post_date": "2022-05-25T08:15:13.803000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>1.I trained three models tf_efficientnet_b0_ns, tf_efficientnet_b3_ns and tf_efficientnet_b5_ns using Kaerururu's code.<br>\n2.Then, on the verification set, the threshold values from 0.01 to 0.49 are traversed to find the best threshold value for every category.<br>\n3.Finally, I ensemble the output of the three models, and the average value of the three optimal thresholds is taken as the final inference threshold.</p>\n<p>In addition, many attempts have been made to no effect, which has been mentioned in other people's public solutions. I'm still a rookie at the competition.</p>",
  "messages": [
    {
      "id": 1800791,
      "postDate": "2022-05-25T08:15:13.803Z",
      "content": "<p>1.I trained three models tf_efficientnet_b0_ns, tf_efficientnet_b3_ns and tf_efficientnet_b5_ns using Kaerururu's code.<br>\n2.Then, on the verification set, the threshold values from 0.01 to 0.49 are traversed to find the best threshold value for every category.<br>\n3.Finally, I ensemble the output of the three models, and the average value of the three optimal thresholds is taken as the final inference threshold.</p>\n<p>In addition, many attempts have been made to no effect, which has been mentioned in other people's public solutions. I'm still a rookie at the competition.</p>",
      "rawMarkdown": "1.I trained three models tf_efficientnet_b0_ns, tf_efficientnet_b3_ns and tf_efficientnet_b5_ns using Kaerururu's code.\n2.Then, on the verification set, the threshold values from 0.01 to 0.49 are traversed to find the best threshold value for every category.\n3.Finally, I ensemble the output of the three models, and the average value of the three optimal thresholds is taken as the final inference threshold.\n\nIn addition, many attempts have been made to no effect, which has been mentioned in other people's public solutions. I'm still a rookie at the competition.",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1800791": "1.I trained three models tf_efficientnet_b0_ns, tf_efficientnet_b3_ns and tf_efficientnet_b5_ns using Kaerururu's code.\n2.Then, on the verification set, the threshold values from 0.01 to 0.49 are traversed to find the best threshold value for every category.\n3.Finally, I ensemble the output of the three models, and the average value of the three optimal thresholds is taken as the final inference threshold.\n\nIn addition, many attempts have been made to no effect, which has been mentioned in other people's public solutions. I'm still a rookie at the competition."
  }
}