{
  "fetched_at": "2026-10-25T04:18:53Z",
  "source_url": "https://www.kaggle.com/competitions/dl-enigma-10-sust-cse-carnival-2024/data",
  "pages": [
    {
      "name": "data-description",
      "content": "Dataset Paper Link: https://arxiv.org/abs/2401.10659\n\n## **Dataset Overview:**\nThe dataset covers the following 9 districts in Bangladesh: Sylhet, Dhaka, Rajshahi, Mymensingh, Maowa, Chittagong, Sirajganj, Sherpur, and Khulna. This dataset contains 9825 images with 78,943 Objects with 13 different classes. It is annotated with rectangular bounding box.Participants will encounter a wide range of road types, including towns, expressways, highways, and village roads. This diversity in locations aims to challenge algorithms to perform well across various driving contexts commonly encountered on Bangladesh roads. \n\n## **Data Collection:**\nTo ensure the authenticity of the dataset, all images were collected using smartphone cameras, simulating the real-world conditions faced by autonomous vehicles. The absence of online images ensures that participants are working with a dataset representative of the actual driving scenarios in Bangladesh, promoting the development of models with practical applicability.\n\n**Goal**: Participants will develop solutions to detect objects in various Road Conditions fo Bangladesh.\n\n## **Classes** : 13\n- auto_rickshaw\n- bicycle\n- bus\n- car\n- cart_vehicle\n- construction_vehicle\n- motorbike\n- person\n- priority_vehicle\n- three_wheeler\n- train\n- truck\n- wheelchair\n\n## **Sample Dataset**:\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F5328059%2F2c69790db325e71f9bfa40fb8d3012d0%2Fimage%20(1).png?generation=1705692179963469&alt=media)\n\n## **Dataset Format**\n\nThis dataset is in Yolov5 format. The train and test images can be found in in `dlenigma1/BadODD/images/`, and the labels for the train dataset can be found in `dlenigma1/BadODD/labels/` directory.\n\n- Train Image Count: 5896\n- Test Image Count: 1964\n\n**Each training image has a txt file of the same name in the labels directory.**\n\n**Test**\nTest set images are in `dlenigma1/BadODD/images/test`.\n\nThe test set has no provided annotations. Your task is to learn the bounding box from the train set and predict it for the test set images.\n\n\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "**This Competition is open to All from any institution in Bangladesh.**\n\nStudent from the **SUST CSE 18** batch is not allowed to participate in this competition.\n\n## **How to participate**\nAnyone can join the kaggle competition for Phase 1. However, to be eligible to participate in Phase 2 and Prizes a team must fill out the Registration Form.\n\n**Registration Link**: https://csecarnival.sust.edu.bd/dlenigma/registration\n**FB Event Link**: https://fb.me/e/vepwR8e5e\n\n## **Competition Schedule**\n\n**Phase 1**: \nDuration: 20 Jan - 10 Feb 2024.\nParticipants will submit their solutions on Kaggle. The top 10 teams will be invited for Phase 2.\n\n**Phase 2**:\nDate: 18 February\nParticipants will have to submit an IEEE format paper and slide to present their work on-site. More details regarding the onsite presentation will be given to the eligible participants for Phase 2.\n\nAll teams will submit working Kaggle notebooks for Train and Test with appropriate instructions and documentation for reproducibility after Phase 1. Test notebook must run on any accelerator of Kaggle under the Kaggle Runtime.\n\n## Scoring\n\nPhase 1: (50%)\n- Public: 30% mAP score\n- Private: 20% mAP score\n\nPhase 2: (50%)\n - Private Test Data: 25% mAP score\n - Judge + Presentation + Paper : 15% \n - Inference time: 10%\n\n*For any inquiries, please feel free to contact us via email at `sustdlenigma@gmail.com`*",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "For this competition, we are using mAP ( Mean Average Precision) as our evaluation metric.\n\nThe mAP is computed by first calculating the Average Precision (AP) for each class and then taking the mean across all classes. AP is determined by computing the area under the precision-recall curve, where precision is plotted against recall for various confidence thresholds.\n\n**For this competition we are evaluating mAP50 where IoU is set to 0.5 rather than calculating the mean of different sets of IoU thresholds.**\n## Submission File\n\nYou will submit a submission csv file where each row of the csv file will contain one or multiple Predicted Bounding Box for each Image. Multiple bounding boxes will be comma separated for one image id. `submission.csv` will contain 3 columns: `id`, `ImageID`, and `PredictionString_pred`.\n\n`id`: This column will contain the row number that will start from 0. \n`ImageID`: This column will be composed of image_id which is the name of the image file without the extension(.jpg).\n`PredictionString_pred`: This will contain Bounding Boxes for multiple objects for each ImageID. Each Bounding Box will be comma(`,`) separated and the format is `class_id confidence_score xmin ymin width height`. Each of these values are floating value.\n\nIf the model predicts no prediction or `NaN` for one ImageID return `0 0 0 0 0 0`. `NaN` values in `submission.csv` will result in a submission error.\n\n**Each value in the PredictionString_pred is normalized.**\n\nIn this image x1, y1 denotes xmin, ymin.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F5328059%2Fd200a198d4e72532b928fe2229c7e819%2FGroup%208%20(1).png?generation=1705755050400160&alt=media)\n\n\n    id, ImageID, PredictionString_pred\n    0, image1, 3.0 1.0 0.3484375 0.4222222222222222 0.11510416666666666 0.08888888888888889\n    1, image2, 4.0 0.7 0.6484375 0.2222222222222222 0.21510416666666666 0.18888888888888889,10.0 0.5 0.9484375 0.1222222222222222 0.31510416666666666 0.48888888888888889\n  \n    etc",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "**Registration is mandatory for all participants to be eligible for prizes**. Registration Link: https://csecarnival.sust.edu.bd/dlenigma/registration\n\n - Only one account per participant in Kaggle will be allowed.\n - All teams must have unique participants and no participant can be a member of multiple teams.\n - Use of external datasets to train the model is prohibited.\n - Publicly available pre-training models are allowed, and the participants must disclose the data used for pre-training the models. Additionally, these model weights must be incorporated into the solution notebook and links should be provided to their sources. Pretraining Models on other datasets are prohibited.\n - All teams will submit working Kaggle notebooks to train and infer with appropriate instructions and documentation for reproducibility during the competition period. Inference notebooks must run on Kaggle (with any accelerator). Lack of reproducibility may lead to disqualification, even for the best solution.\n - Participants must disclose their solution (Train and Test notebooks, Trained Model Weights, Solution Writeup)  after the competition to be eligible for Winning Prizes and Completion Certificates.\n - Participants must refrain from any form of plagiarism or unethical practices. All work must be original, and proper credit must be given to external sources if used.\n - Internet can be used in Notebooks for library installation and external file download. But API calls are strictly prohibited.\n - Manually labeling the test dataset will result in immediate disqualification.\n - All the participants should be currently studying or graduated from any university in Bangladesh.\n - In the event of a change in the evaluation metric, the leaderboard is subject to change.\n\n**Team Guidelines**\n - The maximum team size is 4 members.\n - Within a team, a maximum of 1 individual who has completed their undergraduate studies is permitted.\n - Team members are allowed to be from different universities.\n - For participants from overseas, a minimum of 1 person must be present on-site during the final Presentation.\n - Teams including participants from SUST CSE 18 Batch will be ineligible for Phase 2 and participation from SUST CSE 18 in Phase 1 is strongly discouraged.\n - Private code sharing between teams is not allowed.\n - Teams can merge only in Phase 1. This can only be done before Registration. After registration, no team can change their team name or members.\n - Teams eligible for Phase 2 will be required to write a paper following the IEEE format of a maximum of 4 pages prepare a presentation slide showing their approach and submit it before the Phase 2 presentation.\n - To be eligible for the SUST Rising Team all the participants of a team should be from current students of SUST.\n\n**General Rules**\n - The participants must only use open source code licensed under an Open Source Initiative-approved license (see www.opensource.org) that in no event limits commercial use of such code or model containing or depending on such code.\n - Winner license: The winners will grant the competition hosts and sponsors the following license:\nYou hereby license and will license your winning submission and the source code used to generate the submission under an Open Source Initiative-approved license (see www.opensource.org) that in no event limits commercial use of such code or model containing or depending on such code. For generally commercially available software that you used to generate the submission that is not owned by you but that can be procured by the Competition Sponsor without undue expense, you do not need to grant the license in the preceding sentence for that software. You must also make your solution public.\n - The Competition hosts and sponsors and its affiliates may use your name and likeness for advertising and promotional purposes without additional compensation unless prohibited by law.\n - You warrant that your submission is your original work and, as such, you are the sole and exclusive owner and rights holder of the submission, and you have the right to make the submission and grant all required licenses. To the maximum extent permitted by law, you indemnify and agree to keep indemnified Competition Entities at all times from and against any liability, claims, demands, losses, damages, costs, and expenses resulting from any of your acts, defaults or omissions and/or a breach of any warranty set forth herein.\n - The DL Enigma Organizing Committee reserves the right to cancel or postpone the event for any reason.\n - The organizers serve the right to disqualify any team.\n - Winners will be selected at the sole discretion of the DL Enigma Organizing Committee. All participants must conform to the terms and conditions. The DL Enigma Organizing Committee’s decisions in this regard will be final, and no queries or appeals thereof will be entertained.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "abstract",
      "content": "\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F5328059%2F9b3067b3ff9d867fd242ccb1b6bd9634%2Fsdasdasdas.jpg?generation=1705692279083086&alt=media)\n\nWelcome to the **Autonomous Vehicles for Bangladesh Roads Kaggle competition**!🚗 In this exciting challenge, participants are invited to tackle the unique and complex task of object detection in diverse driving environments across 9 districts in Bangladesh. **SUST CSE Carnival** presents a Deep Learning Event **DL Enigma 1.0** where the goal is to advance the Autonomous Driving Technology for Bangladesh where participants can build unique approaches to develop solutions that can be used for Autonomous Navigation in Self Driving Cars. **BadODD: Bangladeshi Autonomous Driving Object Detection Dataset** is a diverse dataset for Object Detection on Bangladesh Roads that the participants will explore during the competition period.\n\n\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}