{
  "id": 220653,
  "title": "Takeaways from the competition",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/220653",
  "author_name": "SuryaJR_Rafl",
  "post_date": "2021-02-19T05:29:16.147000",
  "votes": 7,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>This is my first serious competition and I had a great learning curve. Thanks to every kaggler who was generous in sharing great notebooks highlighting concepts. Specifically, I would like to mention some of the notebooks which I found useful.</p>\n<ol>\n<li><p><a href=\"https://www.kaggle.com/abhinand05/vision-transformer-vit-tutorial-baseline\" target=\"_blank\">Vision Transformer (ViT): Tutorial + Baseline</a> by <a href=\"https://www.kaggle.com/abhinand05\" target=\"_blank\">@abhinand05</a> - Vision transformer concept explanation and TPU usage </p></li>\n<li><p><a href=\"https://www.kaggle.com/piantic/vision-transformer-vit-visualize-attention-map\" target=\"_blank\">Vision Transformer (ViT) : Visualize Attention Map</a> by <a href=\"https://www.kaggle.com/piantic\" target=\"_blank\">@piantic</a> analysing attention maps </p></li>\n<li><p><a href=\"https://www.kaggle.com/yasufuminakama/cassava-resnext50-32x4d-starter-training\" target=\"_blank\">Cassava / resnext50_32x4d starter [training]</a> by <a href=\"https://www.kaggle.com/yasufuminakama\" target=\"_blank\">@yasufuminakama</a> - resnext50 usage</p></li>\n<li><p><a href=\"https://www.kaggle.com/khyeh0719/pytorch-efficientnet-baseline-train-amp-aug\" target=\"_blank\">(Pytorch Efficientnet Baseline [Train] AMP+Aug</a> by <a href=\"https://www.kaggle.com/khyeh0719\" target=\"_blank\">@khyeh0719</a> - fmix augumentation usage</p></li>\n<li><p><a href=\"https://www.kaggle.com/hengck23/notebook6262cd4cd3\" target=\"_blank\">notebook6262cd4cd3</a> by <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> - Inference reference</p></li>\n<li><p><a href=\"https://www.kaggle.com/cdeotte/forward-selection-oof-ensemble-0-942-private\" target=\"_blank\">Forward Selection OOF Ensemble - [0.942 Private]</a> by <a href=\"https://www.kaggle.com/cdeotte\" target=\"_blank\">@cdeotte</a> </p></li>\n<li><p><a href=\"https://www.kaggle.com/c/plant-pathology-2020-fgvc7/discussion/154056\" target=\"_blank\">Plant pathology 1st solution</a></p></li>\n<li><p><a href=\"https://www.kaggle.com/piantic/learning-with-noisy-labels-pytorch-xla-tpu\" target=\"_blank\">Learning with Noisy Labels - Pytorch XLA(TPU)🔥</a> by <a href=\"https://www.kaggle.com/piantic\" target=\"_blank\">@piantic</a></p></li>\n<li><p><a href=\"https://www.kaggle.com/c/flower-classification-with-tpus/discussion/132191\" target=\"_blank\">How To - Rotation Augmentation GPU/TPU</a> by <a href=\"https://www.kaggle.com/cdeotte\" target=\"_blank\">@cdeotte</a> </p></li>\n</ol>\n<p>Looking back at the competition, I would like to improve on:</p>\n<ol>\n<li>Setting up a standard pipeline for cv projects</li>\n<li>Hyperparameter optimization </li>\n<li>Ensembling methods</li>\n<li>Better TPU usage </li>\n</ol>\n<p>If people can point out to resources relating to above topics, it'd be great. Thanks again for the great experience. Wish you all luck for the upcoming competitions. </p>",
  "messages": [
    {
      "id": 1209913,
      "postDate": "2021-02-19T05:29:16.147Z",
      "content": "<p>Hi,</p>\n<p>This is my first serious competition and I had a great learning curve. Thanks to every kaggler who was generous in sharing great notebooks highlighting concepts. Specifically, I would like to mention some of the notebooks which I found useful.</p>\n<ol>\n<li><p><a href=\"https://www.kaggle.com/abhinand05/vision-transformer-vit-tutorial-baseline\" target=\"_blank\">Vision Transformer (ViT): Tutorial + Baseline</a> by <a href=\"https://www.kaggle.com/abhinand05\" target=\"_blank\">@abhinand05</a> - Vision transformer concept explanation and TPU usage </p></li>\n<li><p><a href=\"https://www.kaggle.com/piantic/vision-transformer-vit-visualize-attention-map\" target=\"_blank\">Vision Transformer (ViT) : Visualize Attention Map</a> by <a href=\"https://www.kaggle.com/piantic\" target=\"_blank\">@piantic</a> analysing attention maps </p></li>\n<li><p><a href=\"https://www.kaggle.com/yasufuminakama/cassava-resnext50-32x4d-starter-training\" target=\"_blank\">Cassava / resnext50_32x4d starter [training]</a> by <a href=\"https://www.kaggle.com/yasufuminakama\" target=\"_blank\">@yasufuminakama</a> - resnext50 usage</p></li>\n<li><p><a href=\"https://www.kaggle.com/khyeh0719/pytorch-efficientnet-baseline-train-amp-aug\" target=\"_blank\">(Pytorch Efficientnet Baseline [Train] AMP+Aug</a> by <a href=\"https://www.kaggle.com/khyeh0719\" target=\"_blank\">@khyeh0719</a> - fmix augumentation usage</p></li>\n<li><p><a href=\"https://www.kaggle.com/hengck23/notebook6262cd4cd3\" target=\"_blank\">notebook6262cd4cd3</a> by <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> - Inference reference</p></li>\n<li><p><a href=\"https://www.kaggle.com/cdeotte/forward-selection-oof-ensemble-0-942-private\" target=\"_blank\">Forward Selection OOF Ensemble - [0.942 Private]</a> by <a href=\"https://www.kaggle.com/cdeotte\" target=\"_blank\">@cdeotte</a> </p></li>\n<li><p><a href=\"https://www.kaggle.com/c/plant-pathology-2020-fgvc7/discussion/154056\" target=\"_blank\">Plant pathology 1st solution</a></p></li>\n<li><p><a href=\"https://www.kaggle.com/piantic/learning-with-noisy-labels-pytorch-xla-tpu\" target=\"_blank\">Learning with Noisy Labels - Pytorch XLA(TPU)🔥</a> by <a href=\"https://www.kaggle.com/piantic\" target=\"_blank\">@piantic</a></p></li>\n<li><p><a href=\"https://www.kaggle.com/c/flower-classification-with-tpus/discussion/132191\" target=\"_blank\">How To - Rotation Augmentation GPU/TPU</a> by <a href=\"https://www.kaggle.com/cdeotte\" target=\"_blank\">@cdeotte</a> </p></li>\n</ol>\n<p>Looking back at the competition, I would like to improve on:</p>\n<ol>\n<li>Setting up a standard pipeline for cv projects</li>\n<li>Hyperparameter optimization </li>\n<li>Ensembling methods</li>\n<li>Better TPU usage </li>\n</ol>\n<p>If people can point out to resources relating to above topics, it'd be great. Thanks again for the great experience. Wish you all luck for the upcoming competitions. </p>",
      "rawMarkdown": "Hi,\n\nThis is my first serious competition and I had a great learning curve. Thanks to every kaggler who was generous in sharing great notebooks highlighting concepts. Specifically, I would like to mention some of the notebooks which I found useful.\n\n1.  [Vision Transformer (ViT): Tutorial + Baseline](https://www.kaggle.com/abhinand05/vision-transformer-vit-tutorial-baseline) by @abhinand05 - Vision transformer concept explanation and TPU usage \n\n2. [Vision Transformer (ViT) : Visualize Attention Map](https://www.kaggle.com/piantic/vision-transformer-vit-visualize-attention-map) by @piantic analysing attention maps \n\n3. [Cassava / resnext50_32x4d starter [training]](https://www.kaggle.com/yasufuminakama/cassava-resnext50-32x4d-starter-training) by @yasufuminakama - resnext50 usage\n\n4. [(Pytorch Efficientnet Baseline [Train] AMP+Aug](https://www.kaggle.com/khyeh0719/pytorch-efficientnet-baseline-train-amp-aug) by @khyeh0719 - fmix augumentation usage\n\n5. [notebook6262cd4cd3](https://www.kaggle.com/hengck23/notebook6262cd4cd3) by @hengck23 - Inference reference\n\n6. [Forward Selection OOF Ensemble - [0.942 Private]](https://www.kaggle.com/cdeotte/forward-selection-oof-ensemble-0-942-private) by @cdeotte \n\n7. [Plant pathology 1st solution](https://www.kaggle.com/c/plant-pathology-2020-fgvc7/discussion/154056)\n\n8. [Learning with Noisy Labels - Pytorch XLA(TPU)🔥](https://www.kaggle.com/piantic/learning-with-noisy-labels-pytorch-xla-tpu) by @piantic\n\n9. [How To - Rotation Augmentation GPU/TPU](https://www.kaggle.com/c/flower-classification-with-tpus/discussion/132191) by @cdeotte \n\n\nLooking back at the competition, I would like to improve on:\n1. Setting up a standard pipeline for cv projects\n2. Hyperparameter optimization \n3. Ensembling methods\n4. Better TPU usage \n\nIf people can point out to resources relating to above topics, it'd be great. Thanks again for the great experience. Wish you all luck for the upcoming competitions. \n\n\n\n\n\n\n\n",
      "votes": 7
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1209913": "Hi,\n\nThis is my first serious competition and I had a great learning curve. Thanks to every kaggler who was generous in sharing great notebooks highlighting concepts. Specifically, I would like to mention some of the notebooks which I found useful.\n\n1.  [Vision Transformer (ViT): Tutorial + Baseline](https://www.kaggle.com/abhinand05/vision-transformer-vit-tutorial-baseline) by @abhinand05 - Vision transformer concept explanation and TPU usage \n\n2. [Vision Transformer (ViT) : Visualize Attention Map](https://www.kaggle.com/piantic/vision-transformer-vit-visualize-attention-map) by @piantic analysing attention maps \n\n3. [Cassava / resnext50_32x4d starter [training]](https://www.kaggle.com/yasufuminakama/cassava-resnext50-32x4d-starter-training) by @yasufuminakama - resnext50 usage\n\n4. [(Pytorch Efficientnet Baseline [Train] AMP+Aug](https://www.kaggle.com/khyeh0719/pytorch-efficientnet-baseline-train-amp-aug) by @khyeh0719 - fmix augumentation usage\n\n5. [notebook6262cd4cd3](https://www.kaggle.com/hengck23/notebook6262cd4cd3) by @hengck23 - Inference reference\n\n6. [Forward Selection OOF Ensemble - [0.942 Private]](https://www.kaggle.com/cdeotte/forward-selection-oof-ensemble-0-942-private) by @cdeotte \n\n7. [Plant pathology 1st solution](https://www.kaggle.com/c/plant-pathology-2020-fgvc7/discussion/154056)\n\n8. [Learning with Noisy Labels - Pytorch XLA(TPU)🔥](https://www.kaggle.com/piantic/learning-with-noisy-labels-pytorch-xla-tpu) by @piantic\n\n9. [How To - Rotation Augmentation GPU/TPU](https://www.kaggle.com/c/flower-classification-with-tpus/discussion/132191) by @cdeotte \n\n\nLooking back at the competition, I would like to improve on:\n1. Setting up a standard pipeline for cv projects\n2. Hyperparameter optimization \n3. Ensembling methods\n4. Better TPU usage \n\nIf people can point out to resources relating to above topics, it'd be great. Thanks again for the great experience. Wish you all luck for the upcoming competitions. \n\n\n\n\n\n\n\n"
  }
}