{
"id": 412824,
"title": "Fast Starting Resources",
"url": "/competitions/vesuvius-challenge-ink-detection/discussion/412824",
"author_name": "Gaju Ahmed",
"post_date": "2023-05-25T12:32:43.470000",
"votes": 6,
"comment_count": 0,
"views": 0,
"content": "Understand the Data
\n\nModel
\n\nOther Info
\n\n
\n
",
"messages": [
{
"id": 2273809,
"postDate": "2023-05-25T12:32:43.470Z",
"content": "Understand the Data
\n\nModel
\n\nOther Info
\n\n
\n
",
"rawMarkdown": "## Understand the Data\n- Welcome from the Vesuvius Challenge team!\n- I have made my own carbonized papyrus\n- [lb0.68-one-fold-fragment_id_1 !!!??? ] my experimental results ... the trick of getting good results?\n- Analogizing Ink detection to other problem domains\n- [EDA] 🌋VC-📜ID ~ Volume layers animation\n- [EDA] A slice-by-slice analysis\n- Improving performance with L1/Hessian denoising\n- Vesuvius Challenge GIF Animation\n- Color analysis: Ink = Ink + Noise. ROI\n- Run Length Encoding Explained\n- 📜3D X-ray scan layers animation\n- ink-id repo\n- I found an error in ink_label\n\n\n## Model \n- 2.5d segmentaion baseline [training]\n- 2.5d segmentaion baseline [inference\n- 2.5d segmentaion model with rotate TTA\n- 3D ResNet baseline [inference]\n- Pytorch UNet baseline (with train code)\n- A simple, high-performance tf.data pipeline\n- Vesuvis PyTorch⚡ MONAI\n- Confirmation of prohibited pre-trained models\n\n## Other Info\n- CV Strategy and choosing confidence threshold?\n- Loss functions for image segmentation\n- Modified DICE Coefficient Implementation\n- My experimental results, which channels you need?\n- SAM applied to Vesuvius Challenge?\n\n\n\n\n",
"votes": 6
}
],
"comments": [],
"raw_markdown_by_id": {
"2273809": "## Understand the Data\n- Welcome from the Vesuvius Challenge team!\n- I have made my own carbonized papyrus\n- [lb0.68-one-fold-fragment_id_1 !!!??? ] my experimental results ... the trick of getting good results?\n- Analogizing Ink detection to other problem domains\n- [EDA] 🌋VC-📜ID ~ Volume layers animation\n- [EDA] A slice-by-slice analysis\n- Improving performance with L1/Hessian denoising\n- Vesuvius Challenge GIF Animation\n- Color analysis: Ink = Ink + Noise. ROI\n- Run Length Encoding Explained\n- 📜3D X-ray scan layers animation\n- ink-id repo\n- I found an error in ink_label\n\n\n## Model \n- 2.5d segmentaion baseline [training]\n- 2.5d segmentaion baseline [inference\n- 2.5d segmentaion model with rotate TTA\n- 3D ResNet baseline [inference]\n- Pytorch UNet baseline (with train code)\n- A simple, high-performance tf.data pipeline\n- Vesuvis PyTorch⚡ MONAI\n- Confirmation of prohibited pre-trained models\n\n## Other Info\n- CV Strategy and choosing confidence threshold?\n- Loss functions for image segmentation\n- Modified DICE Coefficient Implementation\n- My experimental results, which channels you need?\n- SAM applied to Vesuvius Challenge?\n\n\n\n\n"
}
}