{
  "id": 41794,
  "title": "tflearn ",
  "url": "/competitions/dog-breed-identification/discussion/41794",
  "author_name": "",
  "post_date": "2017-10-23T23:59:57.243487100Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Has anybody tried to do it with tflearn? if you did, what would you recommend keras or tflearn? Also, how can I access the pre-trained model with tflearn? </p>",
  "messages": [
    {
      "id": "234748",
      "postDate": "10/23/2017 23:59:57",
      "content": "<p>Has anybody tried to do it with tflearn? if you did, what would you recommend keras or tflearn? Also, how can I access the pre-trained model with tflearn? </p>",
      "rawMarkdown": "Has anybody tried to do it with tflearn? if you did, what would you recommend keras or tflearn? Also, how can I access the pre-trained model with tflearn?",
      "votes": null
    },
    {
      "id": "250866",
      "postDate": "11/30/2017 12:29:06",
      "content": "<p>I tried doing it with TFLearn, my notebook is in the Kernels section of the competition. I couldn't tell you whether TensorFlow/TFLearn or Keras is better, but using the pre-trained model in Keras would be best for a newcomer, whereas building a network/estimator from scratch in TFLearn/TensorFlow/Keras might be for more advanced competitors as the high number of target values in this competition (120 dog breeds!) makes it quite the challenging task to get a high accuracy with the available data set. As to how to load the pre-trained network, I'm not sure how to do it in TFLearn (assuming there is one made for TensorFlow/TFLearn) but for Keras, there are a few kernels out there that explain how to do it.\nHope this helped! \nCheckout my Kernel if you're thinking of using TFLearn to build it from scratch!\nGood luck!</p>",
      "rawMarkdown": "I tried doing it with TFLearn, my notebook is in the Kernels section of the competition. I couldn't tell you whether TensorFlow/TFLearn or Keras is better, but using the pre-trained model in Keras would be best for a newcomer, whereas building a network/estimator from scratch in TFLearn/TensorFlow/Keras might be for more advanced competitors as the high number of target values in this competition (120 dog breeds!) makes it quite the challenging task to get a high accuracy with the available data set. As to how to load the pre-trained network, I'm not sure how to do it in TFLearn (assuming there is one made for TensorFlow/TFLearn) but for Keras, there are a few kernels out there that explain how to do it.\nHope this helped! \nCheckout my Kernel if you're thinking of using TFLearn to build it from scratch!\nGood luck!",
      "votes": null
    },
    {
      "id": "252496",
      "postDate": "12/03/2017 02:59:05",
      "content": "<p>keras has much more excited functions than tflearn.</p>",
      "rawMarkdown": "keras has much more excited functions than tflearn.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 250866,
      "author_name": "rkasincdrk",
      "author_url": "",
      "post_date": "11/30/2017 12:29:06",
      "content": "<p>I tried doing it with TFLearn, my notebook is in the Kernels section of the competition. I couldn't tell you whether TensorFlow/TFLearn or Keras is better, but using the pre-trained model in Keras would be best for a newcomer, whereas building a network/estimator from scratch in TFLearn/TensorFlow/Keras might be for more advanced competitors as the high number of target values in this competition (120 dog breeds!) makes it quite the challenging task to get a high accuracy with the available data set. As to how to load the pre-trained network, I'm not sure how to do it in TFLearn (assuming there is one made for TensorFlow/TFLearn) but for Keras, there are a few kernels out there that explain how to do it.\nHope this helped! \nCheckout my Kernel if you're thinking of using TFLearn to build it from scratch!\nGood luck!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 252496,
      "author_name": "xieenze",
      "author_url": "",
      "post_date": "12/03/2017 02:59:05",
      "content": "<p>keras has much more excited functions than tflearn.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "234748": "Has anybody tried to do it with tflearn? if you did, what would you recommend keras or tflearn? Also, how can I access the pre-trained model with tflearn?",
    "250866": "I tried doing it with TFLearn, my notebook is in the Kernels section of the competition. I couldn't tell you whether TensorFlow/TFLearn or Keras is better, but using the pre-trained model in Keras would be best for a newcomer, whereas building a network/estimator from scratch in TFLearn/TensorFlow/Keras might be for more advanced competitors as the high number of target values in this competition (120 dog breeds!) makes it quite the challenging task to get a high accuracy with the available data set. As to how to load the pre-trained network, I'm not sure how to do it in TFLearn (assuming there is one made for TensorFlow/TFLearn) but for Keras, there are a few kernels out there that explain how to do it.\nHope this helped! \nCheckout my Kernel if you're thinking of using TFLearn to build it from scratch!\nGood luck!",
    "252496": "keras has much more excited functions than tflearn."
  },
  "source": "meta"
}