data.md

Data

This page introduces how to download our dataset and how to use it.

Download

Sensor Data

The files mentioned below can also be downloaded via OpenDataLab. It is recommended to use provided command line interface for acceleration. Note: you only need to download and unzip OpenLane-V2_subset_A_image_*.tar for this track.

Split	Google Drive	Baidu Yun	md5	Size
train	image_0	image_0	8ade7daeec1b64f8ab91a50c81d812f6	~14.0G
	image_1	image_1	c78e776f79e2394d2d5d95b7b5985e0f	~14.3G
	image_2	image_2	4bf09079144aa54cb4dcd5ff6e00cf79	~14.2G
	image_3	image_3	fd9e64345445975f462213b209632aee	~14.4G
	image_4	image_4	ae07e48c88ea2c3f6afbdf5ff71e9821	~14.5G
	image_5	image_5	df62c1f6e6b3fb2a2a0868c78ab19c92	~14.2G
	image_6	image_6	7bff1ce30329235f8e0f25f6f6653b8f	~14.4G
val	image_7	image_7	c73af4a7aef2692b96e4e00795120504	~21.0G
test	image_8	image_8	fb2f61e7309e0b48e2697e085a66a259	~21.2G

For files in Google Drive, you can use the following command by replacing [FILE_ID] and [FILE_NAME] accordingly:

wget --load-cookies /tmp/cookies.txt "https://docs.google.com/uc?export=download&confirm=$(wget --quiet --save-cookies /tmp/cookies.txt --keep-session-cookies --no-check-certificate 'https://docs.google.com/uc?export=download&id=[FILE_ID]' -O- | sed -rn 's/.*confirm=([0-9A-Za-z_]+).*/\1\n/p')&id=[FILE_ID]" -O [FILE_NAME]

Annotations

Please download map annotations and unzip it. Store your data with following structure.

└── data
    ├── argoverse2
    |   ├── [segment_id]
    |   |   ├── image
    |   |   |   ├── [camera]
    |   |   |   |   ├── [timestamp].jpg
    |   |   |   |   └── ...
    |   |   |   └── ...
    |   |   
    |   └── ...
    ├── train_annotations.json
    ├── val_annotations.json
    └── test_annotations.json
├── src
├── tools
...

Data Annotation

Here we introduce the format of our data annotations. The range of annotations is 60 x 30 meters in ego-vehicle coordinate system. All polylines will be truncated at this range.

train_annotations {
    segment_id <str>: {                                     -- data by segment_id (by log).
        list [                                              -- samples in a segment forms a list, ordered by time
            frame_data (dict) {                             -- single frame sample dict
                "segment_id": str,                          -- segment_id, unique by segment
                "timestamp": str,                           -- timestamp (or token), unique by sample
                "sensor": cams (dict) {                     -- cams info dict
                    "ring_front_center": img_metas {        -- single img meta info dict
                        'image_path': str,                  -- corresponding image file path, *.jpg
                        "intrinsic": 3x3 matrix,            -- camera intrinsic matrix, 3x3
                        "extrinsic": 4x4 matrix             -- ego-to-camera transformation matrix, 4x4
                    }
                    "ring_front_left":  ...,                -- img meta for all surrouding cameras
                    "ring_front_right": ...,
                    "ring_side_left":   ...,
                    "ring_side_right":  ...,
                    "ring_rear_left":   ...,
                    "ring_rear_right":  ...,
                },
                "annotation": anns (dict) {                 -- annotations for map elements, indexed by categories
                    "ped_crossing": list [                  -- list of ped crossing lines
                        line1 (N1 x 4),                     -- line as list of points, each point as (x, y, z, visibility)
                                                                visibility: 1 for visible at current frame, 0 for occluded
                        line2 (N2 x 4),
                        ...
                    ]
                    "divider": list [...],                  -- list of divider lines
                    "boundary": list [...]                  -- list of boundary lines
                },
                "pose": ego_pose (dict) {                   -- vehicle ego pose
                    "ego2global_translation": 3-d vector,   -- ego2global translation, in (x, y, z)
                    "ego2global_rotation": 3x3 matrix,      -- ego2global rotation matrix
                }
            }
        ]             
    }
}