{
  "id": 168095,
  "title": "Trained Models",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/168095",
  "author_name": "Durgesh",
  "post_date": "2020-07-19T06:36:55.591000",
  "votes": 5,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi Kagglers,\nI trained(and training more in future also) some variant of EffiecientNet on Various  Image Sizes and made them available for all. \n<strong><em>NOTE: these Models are saved in tensorflow</em></strong></p>\n\n<h3>EfficientNetB0 | 128x128</h3>\n\n<p><a href=\"https://www.kaggle.com/orionpax00/efficientnetb0-128x128\">DATASET FOLD 1-5</a> | <a href=\"https://www.kaggle.com/orionpax00/melanoma-detection-single-model-gpu-tpu-yapl?scriptVersionId=38939103\">NOTEBOOK</a>\n* 5 FOLD AVG LB: <strong>0.9020</strong>\n* FOLD 1 LB: <strong>0.8703</strong>\n* FOLD 3 LB: <strong>0.9072</strong></p>\n\n<h3>EfficientNetB0 | 384x384</h3>\n\n<p><a href=\"https://www.kaggle.com/orionpax00/efficientnetb0-384x384-fold-13\">DATASET FOLD 1-3</a> | <a href=\"https://www.kaggle.com/orionpax00/melanoma-detection-single-model-gpu-tpu-yapl?scriptVersionId=38994028\">NOTEBOOK</a></p>\n\n<ul>\n<li>FOLD 1 LB: <strong>0.9111</strong></li>\n<li>FOLD 2 LB: <strong>0.9084</strong></li>\n<li>FOLD 3 LB: <strong>0.9066</strong></li>\n</ul>\n\n<h3>EfficientNetB6 | 384x384</h3>\n\n<p><a href=\"https://www.kaggle.com/orionpax00/effcientnetb6-384x384-fold-1-3\">DATASET FOLD 1-3</a> | <a href=\"https://www.kaggle.com/orionpax00/effcientnetb6384x384fold45\">DATASET FOLD 4-5</a> | <a href=\"https://www.kaggle.com/orionpax00/effiecientnetb6-384x384-fold4-5-slow?scriptVersionId=39003688\">NOTEBOOK</a></p>\n\n<ul>\n<li>FOLD 1 LB: <strong>0.9145</strong></li>\n<li>FOLD 3 LB: <strong>0.9387</strong></li>\n<li>FOLD 4 LB: <strong>0.9114</strong></li>\n</ul>\n\n<h3>Training Strategy</h3>\n\n<ul>\n<li>5 folds - 15 epoch training</li>\n<li>Stratified splits </li>\n<li>LR Scheduler\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3328754%2F6dc6066f4893b2a50870f6f53941baef%2F__results___20_1.png?generation=1595138685888418&amp;alt=media\" alt=\"\"></li>\n</ul>\n\n<p>SAMPLE NOTEBOOK ON HOW TO USE\n<a href=\"https://www.kaggle.com/orionpax00/melanoma-inference/\">https://www.kaggle.com/orionpax00/melanoma-inference/</a></p>",
  "messages": [
    {
      "id": 935205,
      "postDate": "2020-07-19T06:36:55.590Z",
      "content": "<p>Hi Kagglers,\nI trained(and training more in future also) some variant of EffiecientNet on Various  Image Sizes and made them available for all. \n<strong><em>NOTE: these Models are saved in tensorflow</em></strong></p>\n\n<h3>EfficientNetB0 | 128x128</h3>\n\n<p><a href=\"https://www.kaggle.com/orionpax00/efficientnetb0-128x128\">DATASET FOLD 1-5</a> | <a href=\"https://www.kaggle.com/orionpax00/melanoma-detection-single-model-gpu-tpu-yapl?scriptVersionId=38939103\">NOTEBOOK</a>\n* 5 FOLD AVG LB: <strong>0.9020</strong>\n* FOLD 1 LB: <strong>0.8703</strong>\n* FOLD 3 LB: <strong>0.9072</strong></p>\n\n<h3>EfficientNetB0 | 384x384</h3>\n\n<p><a href=\"https://www.kaggle.com/orionpax00/efficientnetb0-384x384-fold-13\">DATASET FOLD 1-3</a> | <a href=\"https://www.kaggle.com/orionpax00/melanoma-detection-single-model-gpu-tpu-yapl?scriptVersionId=38994028\">NOTEBOOK</a></p>\n\n<ul>\n<li>FOLD 1 LB: <strong>0.9111</strong></li>\n<li>FOLD 2 LB: <strong>0.9084</strong></li>\n<li>FOLD 3 LB: <strong>0.9066</strong></li>\n</ul>\n\n<h3>EfficientNetB6 | 384x384</h3>\n\n<p><a href=\"https://www.kaggle.com/orionpax00/effcientnetb6-384x384-fold-1-3\">DATASET FOLD 1-3</a> | <a href=\"https://www.kaggle.com/orionpax00/effcientnetb6384x384fold45\">DATASET FOLD 4-5</a> | <a href=\"https://www.kaggle.com/orionpax00/effiecientnetb6-384x384-fold4-5-slow?scriptVersionId=39003688\">NOTEBOOK</a></p>\n\n<ul>\n<li>FOLD 1 LB: <strong>0.9145</strong></li>\n<li>FOLD 3 LB: <strong>0.9387</strong></li>\n<li>FOLD 4 LB: <strong>0.9114</strong></li>\n</ul>\n\n<h3>Training Strategy</h3>\n\n<ul>\n<li>5 folds - 15 epoch training</li>\n<li>Stratified splits </li>\n<li>LR Scheduler\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3328754%2F6dc6066f4893b2a50870f6f53941baef%2F__results___20_1.png?generation=1595138685888418&amp;alt=media\" alt=\"\"></li>\n</ul>\n\n<p>SAMPLE NOTEBOOK ON HOW TO USE\n<a href=\"https://www.kaggle.com/orionpax00/melanoma-inference/\">https://www.kaggle.com/orionpax00/melanoma-inference/</a></p>",
      "rawMarkdown": "Hi Kagglers,\nI trained(and training more in future also) some variant of EffiecientNet on Various  Image Sizes and made them available for all. \n***NOTE: these Models are saved in tensorflow***\n### EfficientNetB0 | 128x128\n[DATASET FOLD 1-5](https://www.kaggle.com/orionpax00/efficientnetb0-128x128) | [NOTEBOOK](https://www.kaggle.com/orionpax00/melanoma-detection-single-model-gpu-tpu-yapl?scriptVersionId=38939103)\n* 5 FOLD AVG LB: **0.9020**\n* FOLD 1 LB: **0.8703**\n* FOLD 3 LB: **0.9072**\n\n### EfficientNetB0 | 384x384\n[DATASET FOLD 1-3](https://www.kaggle.com/orionpax00/efficientnetb0-384x384-fold-13) | [NOTEBOOK](https://www.kaggle.com/orionpax00/melanoma-detection-single-model-gpu-tpu-yapl?scriptVersionId=38994028)\n\n* FOLD 1 LB: **0.9111**\n* FOLD 2 LB: **0.9084**\n* FOLD 3 LB: **0.9066**\n\n### EfficientNetB6 | 384x384\n[DATASET FOLD 1-3](https://www.kaggle.com/orionpax00/effcientnetb6-384x384-fold-1-3) | [DATASET FOLD 4-5](https://www.kaggle.com/orionpax00/effcientnetb6384x384fold45) | [NOTEBOOK](https://www.kaggle.com/orionpax00/effiecientnetb6-384x384-fold4-5-slow?scriptVersionId=39003688)\n\n* FOLD 1 LB: **0.9145**\n* FOLD 3 LB: **0.9387**\n* FOLD 4 LB: **0.9114**\n\n### Training Strategy\n* 5 folds - 15 epoch training\n* Stratified splits \n* LR Scheduler\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3328754%2F6dc6066f4893b2a50870f6f53941baef%2F__results___20_1.png?generation=1595138685888418&amp;alt=media)\n\n\nSAMPLE NOTEBOOK ON HOW TO USE\nhttps://www.kaggle.com/orionpax00/melanoma-inference/",
      "votes": 5
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "935205": "Hi Kagglers,\nI trained(and training more in future also) some variant of EffiecientNet on Various  Image Sizes and made them available for all. \n***NOTE: these Models are saved in tensorflow***\n### EfficientNetB0 | 128x128\n[DATASET FOLD 1-5](https://www.kaggle.com/orionpax00/efficientnetb0-128x128) | [NOTEBOOK](https://www.kaggle.com/orionpax00/melanoma-detection-single-model-gpu-tpu-yapl?scriptVersionId=38939103)\n* 5 FOLD AVG LB: **0.9020**\n* FOLD 1 LB: **0.8703**\n* FOLD 3 LB: **0.9072**\n\n### EfficientNetB0 | 384x384\n[DATASET FOLD 1-3](https://www.kaggle.com/orionpax00/efficientnetb0-384x384-fold-13) | [NOTEBOOK](https://www.kaggle.com/orionpax00/melanoma-detection-single-model-gpu-tpu-yapl?scriptVersionId=38994028)\n\n* FOLD 1 LB: **0.9111**\n* FOLD 2 LB: **0.9084**\n* FOLD 3 LB: **0.9066**\n\n### EfficientNetB6 | 384x384\n[DATASET FOLD 1-3](https://www.kaggle.com/orionpax00/effcientnetb6-384x384-fold-1-3) | [DATASET FOLD 4-5](https://www.kaggle.com/orionpax00/effcientnetb6384x384fold45) | [NOTEBOOK](https://www.kaggle.com/orionpax00/effiecientnetb6-384x384-fold4-5-slow?scriptVersionId=39003688)\n\n* FOLD 1 LB: **0.9145**\n* FOLD 3 LB: **0.9387**\n* FOLD 4 LB: **0.9114**\n\n### Training Strategy\n* 5 folds - 15 epoch training\n* Stratified splits \n* LR Scheduler\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3328754%2F6dc6066f4893b2a50870f6f53941baef%2F__results___20_1.png?generation=1595138685888418&amp;alt=media)\n\n\nSAMPLE NOTEBOOK ON HOW TO USE\nhttps://www.kaggle.com/orionpax00/melanoma-inference/"
  }
}