{
  "id": 418406,
  "title": "A Getting-Started Note from The Organizers",
  "url": "/competitions/dlsprint2/discussion/418406",
  "author_name": "",
  "post_date": "2023-06-20T12:49:45.481757100Z",
  "votes": 9,
  "comment_count": 9,
  "views": 0,
  "content": "<p>Hi everyone, welcome to DLSprint 2.0! Hope you guys are as excited as we are. We understand that Deep Learning events can be daunting for first-timers, especially because of the plethora of libraries that exist like numpy, pandas, scikit-learn, seaborn, PyTorch, Tensorflow and on and on and on. </p>\n<p>So we're here to provide a little direction for those looking to get started.</p>\n<p><br></p>\n<h3>If you are new to Deep Learning:</h3>\n<ul>\n<li>It might be best to focus on the big picture of just processing data and training models instead of learning the internal details like gradient descent, loss functions or how this or that model works under the hood.</li>\n<li>We are planning three workshops to introduce participants to Deep Learning, the Competition dataset and some of the current best-performing approaches to the competition problem. <strong>The first online workshop will be a practical Introduction to deep learning competitions on Kaggle intended for people who only know Python</strong>. The tentative date is Saturday (June 24th).</li>\n</ul>\n<p><br></p>\n<h3>If you have prior experience in Deep Learning:</h3>\n<ul>\n<li><a href=\"https://arxiv.org/abs/2303.05325\" target=\"_blank\">This</a> research paper about the BaDLAD Dataset from Bengali.AI is a highly recommended read. The paper already includes some data analysis and evaluation of some well-performing models for the task. We believe going through the paper will save you some time.</li>\n<li>Some of the relevant models are R-CNN, Fast-RCNN, Faster-RCNN, Mask-RCNN and Vision Transformers (ViT). <a href=\"https://github.com/TienLungSun/PyTorch-CV-tasks\" target=\"_blank\">This</a> repository goes in-depth into the inner workings of these models.</li>\n<li>We plan to <strong>introduce the dataset and some of the models in the second online workshop</strong>. However, you are encouraged to get a headstart on the topics.</li>\n<li>We will release <strong>a starter notebook</strong> that uses the <a href=\"https://github.com/facebookresearch/detectron2\" target=\"_blank\">Detectron2</a> library very soon. Detectron2 implements Mask RCNN internally and provides some QoL improvements over the original implementations.<br>\n<br><br>\nHappy Kaggling.</li>\n</ul>",
  "messages": [
    {
      "id": "2310540",
      "postDate": "06/20/2023 12:49:45",
      "content": "<p>Hi everyone, welcome to DLSprint 2.0! Hope you guys are as excited as we are. We understand that Deep Learning events can be daunting for first-timers, especially because of the plethora of libraries that exist like numpy, pandas, scikit-learn, seaborn, PyTorch, Tensorflow and on and on and on. </p>\n<p>So we're here to provide a little direction for those looking to get started.</p>\n<p><br></p>\n<h3>If you are new to Deep Learning:</h3>\n<ul>\n<li>It might be best to focus on the big picture of just processing data and training models instead of learning the internal details like gradient descent, loss functions or how this or that model works under the hood.</li>\n<li>We are planning three workshops to introduce participants to Deep Learning, the Competition dataset and some of the current best-performing approaches to the competition problem. <strong>The first online workshop will be a practical Introduction to deep learning competitions on Kaggle intended for people who only know Python</strong>. The tentative date is Saturday (June 24th).</li>\n</ul>\n<p><br></p>\n<h3>If you have prior experience in Deep Learning:</h3>\n<ul>\n<li><a href=\"https://arxiv.org/abs/2303.05325\" target=\"_blank\">This</a> research paper about the BaDLAD Dataset from Bengali.AI is a highly recommended read. The paper already includes some data analysis and evaluation of some well-performing models for the task. We believe going through the paper will save you some time.</li>\n<li>Some of the relevant models are R-CNN, Fast-RCNN, Faster-RCNN, Mask-RCNN and Vision Transformers (ViT). <a href=\"https://github.com/TienLungSun/PyTorch-CV-tasks\" target=\"_blank\">This</a> repository goes in-depth into the inner workings of these models.</li>\n<li>We plan to <strong>introduce the dataset and some of the models in the second online workshop</strong>. However, you are encouraged to get a headstart on the topics.</li>\n<li>We will release <strong>a starter notebook</strong> that uses the <a href=\"https://github.com/facebookresearch/detectron2\" target=\"_blank\">Detectron2</a> library very soon. Detectron2 implements Mask RCNN internally and provides some QoL improvements over the original implementations.<br>\n<br><br>\nHappy Kaggling.</li>\n</ul>",
      "rawMarkdown": "Hi everyone, welcome to DLSprint 2.0! Hope you guys are as excited as we are. We understand that Deep Learning events can be daunting for first-timers, especially because of the plethora of libraries that exist like numpy, pandas, scikit-learn, seaborn, PyTorch, Tensorflow and on and on and on. \n\nSo we're here to provide a little direction for those looking to get started.\n\n<br>\n### If you are new to Deep Learning:\n- It might be best to focus on the big picture of just processing data and training models instead of learning the internal details like gradient descent, loss functions or how this or that model works under the hood.\n- We are planning three workshops to introduce participants to Deep Learning, the Competition dataset and some of the current best-performing approaches to the competition problem. **The first online workshop will be a practical Introduction to deep learning competitions on Kaggle intended for people who only know Python**. The tentative date is Saturday (June 24th).\n\n<br>\n### If you have prior experience in Deep Learning:\n- [This](https://arxiv.org/abs/2303.05325) research paper about the BaDLAD Dataset from Bengali.AI is a highly recommended read. The paper already includes some data analysis and evaluation of some well-performing models for the task. We believe going through the paper will save you some time.\n- Some of the relevant models are R-CNN, Fast-RCNN, Faster-RCNN, Mask-RCNN and Vision Transformers (ViT). [This](https://github.com/TienLungSun/PyTorch-CV-tasks) repository goes in-depth into the inner workings of these models.\n- We plan to **introduce the dataset and some of the models in the second online workshop**. However, you are encouraged to get a headstart on the topics.\n- We will release **a starter notebook** that uses the [Detectron2](https://github.com/facebookresearch/detectron2) library very soon. Detectron2 implements Mask RCNN internally and provides some QoL improvements over the original implementations.\n<br>\nHappy Kaggling.",
      "votes": null
    },
    {
      "id": "2311325",
      "postDate": "06/21/2023 05:28:19",
      "content": "<p>When should we fill up the registration form? After we individually join the competition or after forming the team?</p>",
      "rawMarkdown": "When should we fill up the registration form? After we individually join the competition or after forming the team?",
      "votes": null
    },
    {
      "id": "2311329",
      "postDate": "06/21/2023 05:35:32",
      "content": "<p>Any time before 28th July. The registration is mandatory to participate in the onsite round and be eligible for the prizes.</p>\n<p>You can register individually or as a team. Both are fine. You can even merge two teams after you register. The team name you provide in the Registration form is NOT mandatory and NOT final. You can change team names anytime before 28th July.</p>",
      "rawMarkdown": "Any time before 28th July. The registration is mandatory to participate in the onsite round and be eligible for the prizes.\n\nYou can register individually or as a team. Both are fine. You can even merge two teams after you register. The team name you provide in the Registration form is NOT mandatory and NOT final. You can change team names anytime before 28th July.",
      "votes": null
    },
    {
      "id": "2311535",
      "postDate": "06/21/2023 08:15:57",
      "content": "<p>Okay. Thank you for clarifying.</p>",
      "rawMarkdown": "Okay. Thank you for clarifying.",
      "votes": null
    },
    {
      "id": "2312286",
      "postDate": "06/21/2023 19:42:17",
      "content": "<p>Shouldn't there be a yaml file for the yolo format of the dataset? just asking :)</p>",
      "rawMarkdown": "Shouldn't there be a yaml file for the yolo format of the dataset? just asking :)",
      "votes": null
    },
    {
      "id": "2312297",
      "postDate": "06/21/2023 19:47:42",
      "content": "<p>By our testing on yolov8, no yaml file was required.</p>",
      "rawMarkdown": "By our testing on yolov8, no yaml file was required.",
      "votes": null
    },
    {
      "id": "2312312",
      "postDate": "06/21/2023 19:58:40",
      "content": "<p>if the dataset was annotated directly on roboflow and trained there or trained on a colab file using roboflow api of the dataset then I think no yaml file would have been needed, but in case of training in kaggel i think a yaml file is needed.  (I am no expert, just trying to understand the whole thing , thank you) </p>",
      "rawMarkdown": "if the dataset was annotated directly on roboflow and trained there or trained on a colab file using roboflow api of the dataset then I think no yaml file would have been needed, but in case of training in kaggel i think a yaml file is needed.  (I am no expert, just trying to understand the whole thing , thank you)",
      "votes": null
    },
    {
      "id": "2312323",
      "postDate": "06/21/2023 20:10:19",
      "content": "<p>We trained yolov8 on Kaggle ourselves and are confident in saying that we did not need an yaml file.</p>",
      "rawMarkdown": "We trained yolov8 on Kaggle ourselves and are confident in saying that we did not need an yaml file.",
      "votes": null
    },
    {
      "id": "2312326",
      "postDate": "06/21/2023 20:15:20",
      "content": "<p>thank you for clarifying.   </p>",
      "rawMarkdown": "thank you for clarifying.",
      "votes": null
    },
    {
      "id": "2323365",
      "postDate": "06/29/2023 22:16:19",
      "content": "<p>I am Running out of kaggle GPU. How can I increase GPU time in kaggle for free..  Or there is any other way to getting free GPU? <br>\nThanks in Advance.</p>",
      "rawMarkdown": "I am Running out of kaggle GPU. How can I increase GPU time in kaggle for free..  Or there is any other way to getting free GPU? \nThanks in Advance.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2311325,
      "author_name": "mdalimranabir",
      "author_url": "",
      "post_date": "06/21/2023 05:28:19",
      "content": "<p>When should we fill up the registration form? After we individually join the competition or after forming the team?</p>",
      "votes": null,
      "replies": [
        {
          "id": 2311329,
          "author_name": "sameen53",
          "author_url": "",
          "post_date": "06/21/2023 05:35:32",
          "content": "<p>Any time before 28th July. The registration is mandatory to participate in the onsite round and be eligible for the prizes.</p>\n<p>You can register individually or as a team. Both are fine. You can even merge two teams after you register. The team name you provide in the Registration form is NOT mandatory and NOT final. You can change team names anytime before 28th July.</p>",
          "votes": null,
          "replies": [
            {
              "id": 2311535,
              "author_name": "mdalimranabir",
              "author_url": "",
              "post_date": "06/21/2023 08:15:57",
              "content": "<p>Okay. Thank you for clarifying.</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 2312286,
      "author_name": "riazmahmudsunny",
      "author_url": "",
      "post_date": "06/21/2023 19:42:17",
      "content": "<p>Shouldn't there be a yaml file for the yolo format of the dataset? just asking :)</p>",
      "votes": null,
      "replies": [
        {
          "id": 2312297,
          "author_name": "sameen53",
          "author_url": "",
          "post_date": "06/21/2023 19:47:42",
          "content": "<p>By our testing on yolov8, no yaml file was required.</p>",
          "votes": null,
          "replies": [
            {
              "id": 2312312,
              "author_name": "riazmahmudsunny",
              "author_url": "",
              "post_date": "06/21/2023 19:58:40",
              "content": "<p>if the dataset was annotated directly on roboflow and trained there or trained on a colab file using roboflow api of the dataset then I think no yaml file would have been needed, but in case of training in kaggel i think a yaml file is needed.  (I am no expert, just trying to understand the whole thing , thank you) </p>",
              "votes": null,
              "replies": [
                {
                  "id": 2312323,
                  "author_name": "sameen53",
                  "author_url": "",
                  "post_date": "06/21/2023 20:10:19",
                  "content": "<p>We trained yolov8 on Kaggle ourselves and are confident in saying that we did not need an yaml file.</p>",
                  "votes": null,
                  "replies": [
                    {
                      "id": 2312326,
                      "author_name": "riazmahmudsunny",
                      "author_url": "",
                      "post_date": "06/21/2023 20:15:20",
                      "content": "<p>thank you for clarifying.   </p>",
                      "votes": null,
                      "replies": []
                    }
                  ]
                }
              ]
            }
          ]
        }
      ]
    },
    {
      "id": 2323365,
      "author_name": "samratabduljalil",
      "author_url": "",
      "post_date": "06/29/2023 22:16:19",
      "content": "<p>I am Running out of kaggle GPU. How can I increase GPU time in kaggle for free..  Or there is any other way to getting free GPU? <br>\nThanks in Advance.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2310540": "Hi everyone, welcome to DLSprint 2.0! Hope you guys are as excited as we are. We understand that Deep Learning events can be daunting for first-timers, especially because of the plethora of libraries that exist like numpy, pandas, scikit-learn, seaborn, PyTorch, Tensorflow and on and on and on. \n\nSo we're here to provide a little direction for those looking to get started.\n\n<br>\n### If you are new to Deep Learning:\n- It might be best to focus on the big picture of just processing data and training models instead of learning the internal details like gradient descent, loss functions or how this or that model works under the hood.\n- We are planning three workshops to introduce participants to Deep Learning, the Competition dataset and some of the current best-performing approaches to the competition problem. **The first online workshop will be a practical Introduction to deep learning competitions on Kaggle intended for people who only know Python**. The tentative date is Saturday (June 24th).\n\n<br>\n### If you have prior experience in Deep Learning:\n- [This](https://arxiv.org/abs/2303.05325) research paper about the BaDLAD Dataset from Bengali.AI is a highly recommended read. The paper already includes some data analysis and evaluation of some well-performing models for the task. We believe going through the paper will save you some time.\n- Some of the relevant models are R-CNN, Fast-RCNN, Faster-RCNN, Mask-RCNN and Vision Transformers (ViT). [This](https://github.com/TienLungSun/PyTorch-CV-tasks) repository goes in-depth into the inner workings of these models.\n- We plan to **introduce the dataset and some of the models in the second online workshop**. However, you are encouraged to get a headstart on the topics.\n- We will release **a starter notebook** that uses the [Detectron2](https://github.com/facebookresearch/detectron2) library very soon. Detectron2 implements Mask RCNN internally and provides some QoL improvements over the original implementations.\n<br>\nHappy Kaggling.",
    "2311325": "When should we fill up the registration form? After we individually join the competition or after forming the team?",
    "2311329": "Any time before 28th July. The registration is mandatory to participate in the onsite round and be eligible for the prizes.\n\nYou can register individually or as a team. Both are fine. You can even merge two teams after you register. The team name you provide in the Registration form is NOT mandatory and NOT final. You can change team names anytime before 28th July.",
    "2311535": "Okay. Thank you for clarifying.",
    "2312286": "Shouldn't there be a yaml file for the yolo format of the dataset? just asking :)",
    "2312297": "By our testing on yolov8, no yaml file was required.",
    "2312312": "if the dataset was annotated directly on roboflow and trained there or trained on a colab file using roboflow api of the dataset then I think no yaml file would have been needed, but in case of training in kaggel i think a yaml file is needed.  (I am no expert, just trying to understand the whole thing , thank you)",
    "2312323": "We trained yolov8 on Kaggle ourselves and are confident in saying that we did not need an yaml file.",
    "2312326": "thank you for clarifying.",
    "2323365": "I am Running out of kaggle GPU. How can I increase GPU time in kaggle for free..  Or there is any other way to getting free GPU? \nThanks in Advance."
  },
  "source": "meta"
}