MARS#

Format specification#

MARS is a dataset for the motion analysis and person identification task. MARS dataset is available for downloading here

Supported types of annotations:

Bbox

Required attributes:

person_id (str): four-digit number that represent ID of pedestrian;
camera_id (int): one-digit number that represent ID of camera that took the image (original dataset has totally 6 cameras);
track_id (int): four-digit number that represent ID of the track with the particular pedestrian;
frame_id (int): three-digit number, that mean number of frame within this track. For the tracks, their names are accumulated for each ID, but for frames, they start from “0001” in each track.

Import MARS dataset#

Use these instructions to import MARS dataset into Datumaro project:

datum project create
datum project add -f mars ./dataset

Note: the directory with dataset should be subdirectory of the project directory.

mars_dataset
├── <bbox_subset_name1>
│   ├── 0001 # directory with images of pedestrian with id 0001
│   │   ├── 0001C1T0001F001.jpg
│   │   ├── 0001C1T0001F002.jpg
│   │   ├── ...
│   ├── 0002 # directory with images of pedestrian with id 0002
│   │   ├── 0002C1T0001F001.jpg
│   │   ├── 0002C1T0001F001.jpg
│   │   ├── ...
│   ├── 0000 # distractors images, which negatively affect retrieval accuracy.
│   │   ├── 0000C1T0001F001.jpg
│   │   ├── 0000C1T0001F001.jpg
│   │   ├── ...
│   ├── 00-1 # junk images which do not affect retrieval accuracy
│   │   ├── 00-1C1T0001F001.jpg
│   │   ├── 00-1C1T0001F001.jpg
│   │   ├── ...
├── <bbox_subset_name2>
│   ├── ...
├── ...

All images in MARS dataset follow a strict convention of naming:

xxxxCxTxxxxFxxx.jpg

the first four digits indicate the pedestrian’s number;
digit after C indicates the camera id;
four digits after T indicate the track id for this pedestrian;
three digits after F indicate the frame id with this track.

Note: there are two specific pedestrian IDs 0000 and 00-1 which indicate distracting images and unwanted images respectively.