Datasets:

handsomeYun
/

M2I

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 83, in _split_generators
                  raise ValueError(
                  ...<2 lines>...
                  )
              ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 66, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

M2I: Multi-camera Multi-layout Infrastructure Dataset

The M2I (Multi-camera Multi-layout Infrastructure) dataset is a large-scale synthetic dataset for infrastructure-based multi-camera 3D object detection. It is introduced in our paper:

MIC-BEV: Multi-Infrastructure Camera Bird's-Eye-View Transformer with Relation-Aware Fusion for 3D Object Detection

M2I is designed to support research on intelligent transportation systems by providing diverse infrastructure camera deployments, road layouts, environmental conditions, and traffic scenarios for Bird's-Eye-View (BEV) perception.

Overview

M2I provides synchronized multi-camera observations together with 3D object annotations for infrastructure perception.

The dataset includes

Multi-camera RGB images
Camera calibration (intrinsics and extrinsics)
3D bounding box annotations
BEV semantic maps
Temporal sequences
Camera visibility information

Dataset Features

Compared with existing infrastructure perception datasets, M2I provides:

Diverse road layouts (intersections, highways, roundabouts, ramps, etc.)
Variable camera configurations (1–4 cameras)
Heterogeneous camera intrinsics and extrinsics
Multiple camera mounting locations and heights
Diverse weather conditions
Diverse illumination conditions
Dynamic traffic density
Sensor degradation (blur, masking, calibration perturbation)

Dataset Structure

After downloading the dataset from Hugging Face, the repository contains:

M2I/

├── M2I_split_dataset.tar.gz.part-*
├── README.md

The RGB images are stored as multipart archives to simplify distribution on Hugging Face.

After extracting the archive, the dataset is organized as

M2I/

├── M2I_split_dataset/
│   ├── train/
│   ├── val/
│   └── test/
├── M2I_json/
├── Town01.xodr
├── Town03.xodr
├── Town04.xodr
├── Town05.xodr
├── Town06.xodr
├── Town07.xodr
├── Town10HD_Opt.xodr
└── README.md

where

M2I_split_dataset/ contains the RGB images and annotations.
M2I_json/ contains the raw frame-level metadata before conversion to training annotations.
Town0X.xodr and Town.xodr* contain the CARLA HD maps used to generate the scenes.

Download

Download the dataset using the Hugging Face CLI:

hf download handsomeYun/M2I \
  --repo-type dataset \
  --local-dir /path/to/M2I

Extract the image archive:

mkdir -p /path/to/M2I_data

cat M2I_split_dataset.tar.gz.part-* | tar -xzvf - -C /path/to/M2I_data

Preparing Training Annotations

After extracting the dataset, generate the training annotations using the MIC-BEV preprocessing script:

python tools/data_converter/mic-bev/create_v2xset_multiple_map.py

Please refer to the MIC-BEV repository for detailed training instructions.

Related Project

The official implementation of MIC-BEV is available at

https://github.com/HandsomeYun/MIC-BEV

The repository contains

Installation instructions
Dataset preprocessing
Training
Evaluation
Visualization

License

The M2I dataset is released under the CC BY 4.0 license.

Please also comply with the licenses of any third-party software or datasets used together with this project.

Acknowledgements

The M2I dataset was developed by the UCLA Mobility Lab as part of the MIC-BEV project.

MIC-BEV builds upon several outstanding open-source projects, including:

BEVFormer
DETR3D
DD3D
MMDetection3D
RoScenes

Citation

If you use M2I in your research, please cite:

@article{zhang2025micbev,
  title={MIC-BEV: Multi-Infrastructure Camera Bird's-Eye-View Transformer with Relation-Aware Fusion for 3D Object Detection},
  author={Zhang, Yun and Zheng, Zhaoliang and Liu, Johnson and Huang, Zhiyu and Zhou, Zewei and Meng, Zonglin and Cai, Tianhui and Ma, Jiaqi},
  journal={arXiv preprint arXiv:2510.24688},
  year={2025}
}

Downloads last month: 438

Total file size:

732 GB

Paper for handsomeYun/M2I

MIC-BEV: Multi-Infrastructure Camera Bird's-Eye-View Transformer with Relation-Aware Fusion for 3D Object Detection

Paper • 2510.24688 • Published Oct 28, 2025