We’re excited to share a new paper detailing the dataset behind the Nexar Dashcam Crash Prediction Challenge! This paper provides an in-depth look at the dataset’s structure, how was the data collected and processed, and the reasoning behind key design choices.
\n📄 Read the paper here: https://arxiv.org/pdf/2503.03848
\nIf you're taking part in the competition, this paper can help you better understand the dataset and refine your approach. Let us know if you have questions - we’d love to hear your thoughts! 🚀
", "messages": [ { "id": "3143962", "postDate": "03/07/2025 18:54:57", "content": "We’re excited to share a new paper detailing the dataset behind the Nexar Dashcam Crash Prediction Challenge! This paper provides an in-depth look at the dataset’s structure, how was the data collected and processed, and the reasoning behind key design choices.
\n📄 Read the paper here: https://arxiv.org/pdf/2503.03848
\nIf you're taking part in the competition, this paper can help you better understand the dataset and refine your approach. Let us know if you have questions - we’d love to hear your thoughts! 🚀
", "rawMarkdown": "We’re excited to share a new paper detailing the dataset behind the Nexar Dashcam Crash Prediction Challenge! This paper provides an in-depth look at the dataset’s structure, how was the data collected and processed, and the reasoning behind key design choices.\n\n📄 **Read the paper here**: https://arxiv.org/pdf/2503.03848\n\nIf you're taking part in the competition, this paper can help you better understand the dataset and refine your approach. Let us know if you have questions - we’d love to hear your thoughts! 🚀", "votes": null }, { "id": "3162564", "postDate": "03/29/2025 11:27:21", "content": "\"For negative examples, fake event times were generated by adding Gaussian noise to half of the duration of
\nthe video, and by generating fake alert times to match the
\ndistribution of the positive cases.\"
Could you expand on what you mean by this?
", "rawMarkdown": "\"For negative examples, fake event times were generated by adding Gaussian noise to half of the duration of\nthe video, and by generating fake alert times to match the\ndistribution of the positive cases.\"\n\nCould you expand on what you mean by this?", "votes": null }, { "id": "3167126", "postDate": "04/01/2025 09:40:25", "content": "When generating the test set, we cut the video right before the accident. In regular cases there is no accident, so we need to fake the time of the accident and the correspondent alert time.
", "rawMarkdown": "When generating the test set, we cut the video right before the accident. In regular cases there is no accident, so we need to fake the time of the accident and the correspondent alert time.", "votes": null }, { "id": "3171334", "postDate": "04/05/2025 15:19:59", "content": "Thank you for the answer, but I was looking for an answer that expanded on what you mean by adding gaussian noise and fake alert times. What exactly entails a fake alert time? Why is gaussian noise injection necessary?
", "rawMarkdown": "Thank you for the answer, but I was looking for an answer that expanded on what you mean by adding gaussian noise and fake alert times. What exactly entails a fake alert time? Why is gaussian noise injection necessary?", "votes": null } ], "comments": [ { "id": 3162564, "author_name": "elhnan", "author_url": "", "post_date": "03/29/2025 11:27:21", "content": "\"For negative examples, fake event times were generated by adding Gaussian noise to half of the duration of
\nthe video, and by generating fake alert times to match the
\ndistribution of the positive cases.\"
Could you expand on what you mean by this?
", "votes": null, "replies": [ { "id": 3167126, "author_name": "danielcmoura", "author_url": "", "post_date": "04/01/2025 09:40:25", "content": "When generating the test set, we cut the video right before the accident. In regular cases there is no accident, so we need to fake the time of the accident and the correspondent alert time.
", "votes": null, "replies": [ { "id": 3171334, "author_name": "elhnan", "author_url": "", "post_date": "04/05/2025 15:19:59", "content": "Thank you for the answer, but I was looking for an answer that expanded on what you mean by adding gaussian noise and fake alert times. What exactly entails a fake alert time? Why is gaussian noise injection necessary?
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3143962": "We’re excited to share a new paper detailing the dataset behind the Nexar Dashcam Crash Prediction Challenge! This paper provides an in-depth look at the dataset’s structure, how was the data collected and processed, and the reasoning behind key design choices.\n\n📄 **Read the paper here**: https://arxiv.org/pdf/2503.03848\n\nIf you're taking part in the competition, this paper can help you better understand the dataset and refine your approach. Let us know if you have questions - we’d love to hear your thoughts! 🚀", "3162564": "\"For negative examples, fake event times were generated by adding Gaussian noise to half of the duration of\nthe video, and by generating fake alert times to match the\ndistribution of the positive cases.\"\n\nCould you expand on what you mean by this?", "3167126": "When generating the test set, we cut the video right before the accident. In regular cases there is no accident, so we need to fake the time of the accident and the correspondent alert time.", "3171334": "Thank you for the answer, but I was looking for an answer that expanded on what you mean by adding gaussian noise and fake alert times. What exactly entails a fake alert time? Why is gaussian noise injection necessary?" }, "source": "meta" }