{
  "id": 247870,
  "title": "EfficientNet B3/B4/B5 🔥 Baseline Metrics Dashboard + Weights Provided",
  "url": "/competitions/seti-breakthrough-listen/discussion/247870",
  "author_name": "Saurav Maheshkar ☕️",
  "post_date": "2021-06-21T13:39:42.387000",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p><img src=\"https://github.com/SauravMaheshkar/SETI-Breakthrough-Listen/blob/main/assets/Github%20Banner.png?raw=true\" alt=\"\"></p>\n<h1>Link to the <a href=\"https://wandb.ai/sauravmaheshkar/seti\" target=\"_blank\"><strong>Weights and Biases Dashboard</strong></a>.</h1>\n<p>The Model Weights for <strong>EfficientNet(B3/B4/B5)</strong> trained on 512 x 512 images is available in the <a href=\"https://www.kaggle.com/sauravmaheshkar/seti-512-efficientnet-b3b4b5\" target=\"_blank\"><strong>SETI - 512 - EfficientNet - B3/B4/B5 Dataset</strong></a>.</p>\n<h1>Results 📋</h1>\n<p><img src=\"https://raw.githubusercontent.com/SauravMaheshkar/SETI-Breakthrough-Listen/8d63f7e0ebaf3a214263ed413c131241704462fd/assets/Baseline-EfficientNet-Loss.svg\" alt=\"\"></p>\n<p>The following table contains the metrics averaged across <strong>5 (GroupKFold) runs for B3 and B4 and 4 for B5</strong> trained for <strong>30 epochs</strong> using <strong>Batch Size of 128</strong> and <strong>Adam optimizer</strong>.</p>\n<table>\n<thead>\n<tr>\n<th>Name</th>\n<th>loss</th>\n<th>val_loss</th>\n<th>best_val_loss</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>EfficientNetB5-512-Baseline</td>\n<td>0.31094212085008</td>\n<td>0.31083238124847</td>\n<td>0.31079388409852</td>\n</tr>\n<tr>\n<td>EfficientNetB4-512-Baseline</td>\n<td>0.31089267134666</td>\n<td>0.31074333190917</td>\n<td>0.31069145202636</td>\n</tr>\n<tr>\n<td>EfficientNetB3-512-Baseline</td>\n<td>0.311157929897308</td>\n<td>0.31072427034378</td>\n<td>0.310709983110427</td>\n</tr>\n</tbody>\n</table>",
  "messages": [
    {
      "id": 1359742,
      "postDate": "2021-06-21T13:39:42.387Z",
      "content": "<p><img src=\"https://github.com/SauravMaheshkar/SETI-Breakthrough-Listen/blob/main/assets/Github%20Banner.png?raw=true\" alt=\"\"></p>\n<h1>Link to the <a href=\"https://wandb.ai/sauravmaheshkar/seti\" target=\"_blank\"><strong>Weights and Biases Dashboard</strong></a>.</h1>\n<p>The Model Weights for <strong>EfficientNet(B3/B4/B5)</strong> trained on 512 x 512 images is available in the <a href=\"https://www.kaggle.com/sauravmaheshkar/seti-512-efficientnet-b3b4b5\" target=\"_blank\"><strong>SETI - 512 - EfficientNet - B3/B4/B5 Dataset</strong></a>.</p>\n<h1>Results 📋</h1>\n<p><img src=\"https://raw.githubusercontent.com/SauravMaheshkar/SETI-Breakthrough-Listen/8d63f7e0ebaf3a214263ed413c131241704462fd/assets/Baseline-EfficientNet-Loss.svg\" alt=\"\"></p>\n<p>The following table contains the metrics averaged across <strong>5 (GroupKFold) runs for B3 and B4 and 4 for B5</strong> trained for <strong>30 epochs</strong> using <strong>Batch Size of 128</strong> and <strong>Adam optimizer</strong>.</p>\n<table>\n<thead>\n<tr>\n<th>Name</th>\n<th>loss</th>\n<th>val_loss</th>\n<th>best_val_loss</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>EfficientNetB5-512-Baseline</td>\n<td>0.31094212085008</td>\n<td>0.31083238124847</td>\n<td>0.31079388409852</td>\n</tr>\n<tr>\n<td>EfficientNetB4-512-Baseline</td>\n<td>0.31089267134666</td>\n<td>0.31074333190917</td>\n<td>0.31069145202636</td>\n</tr>\n<tr>\n<td>EfficientNetB3-512-Baseline</td>\n<td>0.311157929897308</td>\n<td>0.31072427034378</td>\n<td>0.310709983110427</td>\n</tr>\n</tbody>\n</table>",
      "rawMarkdown": "![](https://github.com/SauravMaheshkar/SETI-Breakthrough-Listen/blob/main/assets/Github%20Banner.png?raw=true)\n\n# Link to the [**Weights and Biases Dashboard**](https://wandb.ai/sauravmaheshkar/seti).\n\nThe Model Weights for **EfficientNet(B3/B4/B5)** trained on 512 x 512 images is available in the [**SETI - 512 - EfficientNet - B3/B4/B5 Dataset**](https://www.kaggle.com/sauravmaheshkar/seti-512-efficientnet-b3b4b5).\n\n\n# Results 📋\n\n![](https://raw.githubusercontent.com/SauravMaheshkar/SETI-Breakthrough-Listen/8d63f7e0ebaf3a214263ed413c131241704462fd/assets/Baseline-EfficientNet-Loss.svg)\n\nThe following table contains the metrics averaged across **5 (GroupKFold) runs for B3 and B4 and 4 for B5** trained for **30 epochs** using **Batch Size of 128** and **Adam optimizer**.\n\n|Name                       |loss               |val_loss           |best_val_loss      |\n|---------------------------|-------------------|-------------------|-------------------|\n|EfficientNetB5-512-Baseline|0.31094212085008 |0.31083238124847 |0.31079388409852 |\n|EfficientNetB4-512-Baseline|0.31089267134666|0.31074333190917 |0.31069145202636|\n|EfficientNetB3-512-Baseline|0.311157929897308|0.31072427034378|0.310709983110427|\n\n\n\n\n\n",
      "votes": 4
    },
    {
      "id": 1360274,
      "postDate": "2021-06-22T01:38:31.083Z",
      "content": "<p>Thank you for sharing, Saurav. <br>\nIt will great help for me!</p>",
      "rawMarkdown": "Thank you for sharing, Saurav. \nIt will great help for me!",
      "votes": 1,
      "replies": [
        {
          "id": 1360501,
          "postDate": "2021-06-22T06:33:12.283Z",
          "content": "<p>Thanks for appreciating 😄. I'll be releasing another dashboard comparing ResNet's pretty soon.</p>",
          "rawMarkdown": "Thanks for appreciating 😄. I'll be releasing another dashboard comparing ResNet's pretty soon."
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 1360274,
      "author_name": "WOOSUNG YOON",
      "author_url": "",
      "post_date": "2021-06-22T01:38:31.083000",
      "content": "<p>Thank you for sharing, Saurav. <br>\nIt will great help for me!</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1360501,
          "author_name": "Saurav Maheshkar ☕️",
          "author_url": "",
          "post_date": "2021-06-22T06:33:12.283000",
          "content": "<p>Thanks for appreciating 😄. I'll be releasing another dashboard comparing ResNet's pretty soon.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1359742": "![](https://github.com/SauravMaheshkar/SETI-Breakthrough-Listen/blob/main/assets/Github%20Banner.png?raw=true)\n\n# Link to the [**Weights and Biases Dashboard**](https://wandb.ai/sauravmaheshkar/seti).\n\nThe Model Weights for **EfficientNet(B3/B4/B5)** trained on 512 x 512 images is available in the [**SETI - 512 - EfficientNet - B3/B4/B5 Dataset**](https://www.kaggle.com/sauravmaheshkar/seti-512-efficientnet-b3b4b5).\n\n\n# Results 📋\n\n![](https://raw.githubusercontent.com/SauravMaheshkar/SETI-Breakthrough-Listen/8d63f7e0ebaf3a214263ed413c131241704462fd/assets/Baseline-EfficientNet-Loss.svg)\n\nThe following table contains the metrics averaged across **5 (GroupKFold) runs for B3 and B4 and 4 for B5** trained for **30 epochs** using **Batch Size of 128** and **Adam optimizer**.\n\n|Name                       |loss               |val_loss           |best_val_loss      |\n|---------------------------|-------------------|-------------------|-------------------|\n|EfficientNetB5-512-Baseline|0.31094212085008 |0.31083238124847 |0.31079388409852 |\n|EfficientNetB4-512-Baseline|0.31089267134666|0.31074333190917 |0.31069145202636|\n|EfficientNetB3-512-Baseline|0.311157929897308|0.31072427034378|0.310709983110427|\n\n\n\n\n\n",
    "1360274": "Thank you for sharing, Saurav. \nIt will great help for me!"
  }
}