{
  "id": 74352,
  "title": "Got a problem in keras?",
  "url": "/competitions/humpback-whale-identification/discussion/74352",
  "author_name": "",
  "post_date": "2018-12-11T13:20:44.117211300Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>i train my model with keras by using resnet50 or vgg16 or densenet. When training the model(Adam), i got loss go down but top_5_category_acc is about 0.007(lr = 0.0001 or lower), then i set lr to 0.001 or higher, i got loss go up but top_5_category_acc go up as well and it reached 1.0 at each epoch end. So What happened in my model? How to solve it? Thanks :)</p>",
  "messages": [
    {
      "id": "437161",
      "postDate": "12/11/2018 13:20:44",
      "content": "<p>i train my model with keras by using resnet50 or vgg16 or densenet. When training the model(Adam), i got loss go down but top_5_category_acc is about 0.007(lr = 0.0001 or lower), then i set lr to 0.001 or higher, i got loss go up but top_5_category_acc go up as well and it reached 1.0 at each epoch end. So What happened in my model? How to solve it? Thanks :)</p>",
      "rawMarkdown": "i train my model with keras by using resnet50 or vgg16 or densenet. When training the model(Adam), i got loss go down but top_5_category_acc is about 0.007(lr = 0.0001 or lower), then i set lr to 0.001 or higher, i got loss go up but top_5_category_acc go up as well and it reached 1.0 at each epoch end. So What happened in my model? How to solve it? Thanks :)",
      "votes": null
    },
    {
      "id": "443281",
      "postDate": "12/21/2018 10:51:57",
      "content": "<p>Hi TsungHan,</p>\n\n<p>Unfortunately, I'm also running in the same problem (Using pertained ResNet50 with the last conv block set to trainable along with the dense classifier) and I'm getting terrible results. Did you find anything wrong?</p>",
      "rawMarkdown": "Hi TsungHan,\n\nUnfortunately, I'm also running in the same problem (Using pertained ResNet50 with the last conv block set to trainable along with the dense classifier) and I'm getting terrible results. Did you find anything wrong?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 443281,
      "author_name": "whatvermawhat",
      "author_url": "",
      "post_date": "12/21/2018 10:51:57",
      "content": "<p>Hi TsungHan,</p>\n\n<p>Unfortunately, I'm also running in the same problem (Using pertained ResNet50 with the last conv block set to trainable along with the dense classifier) and I'm getting terrible results. Did you find anything wrong?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "437161": "i train my model with keras by using resnet50 or vgg16 or densenet. When training the model(Adam), i got loss go down but top_5_category_acc is about 0.007(lr = 0.0001 or lower), then i set lr to 0.001 or higher, i got loss go up but top_5_category_acc go up as well and it reached 1.0 at each epoch end. So What happened in my model? How to solve it? Thanks :)",
    "443281": "Hi TsungHan,\n\nUnfortunately, I'm also running in the same problem (Using pertained ResNet50 with the last conv block set to trainable along with the dense classifier) and I'm getting terrible results. Did you find anything wrong?"
  },
  "source": "meta"
}