DSBD-Research/MetaPKLot-Dataset

The MetaPKLot

This is the repository for the of the MetaPKLot dataset.

The PKLot database is licensed under a Creative Commons Attribution 4.0 International License.

The CNPark-EXT database is lincensed under a Open Data Commons Open Database License (ODbL) v1.0. and can be found at (http://cnrpark.it/)

More information about PLds can be found at tools/PLds.

JSON Structure

The JSON files follow the COCO structure, with some changes to accommodate the specific annotations of the datasets. They were divided into two types: Cars and Spots.

MetaPKLot
└── jsons
    ├── cars
    │   ├── PKLot
    │   │   ├── pucpr_cars.json
    │   │   ├── ufpr04_cars.json
    │   │   └── ufpr05_cars.json
    |   ├── CNRPark-EXT
    │   │   ├── cnr-camera-1_cars.json
    │   │   ├── cnr-camera-2_cars.json
    │   │   ├── ...        
    │   │   └── cnr-camera-9_cars.json
    │   └── PLds
    │       ├── isshk_cars.json
    │       ├── qridr_cars.json
    │       ├── vmlix_cars.json
    │       └── vxusd_cars.json
    └── spots
        ├── PKLot
        │   ├── pucpr_spots.json
        │   ├── ufpr04_spots.json
        │   └── ufpr05_spots.json
        ├── CNRPark-EXT
        │   ├── cnr-camera-1_spots.json
        │   ├── cnr-camera-2_spots.json
        │   ├── ...
        │   └── cnr-camera-9_spots.json
        └── PLds
            ├── isshk_spots.json
            ├── qridr_spots.json
            ├── vmlix_spots.json
            └── vxusd_spots.json

Notes of the JSON formats

The annotations_rectangle field define a rectangle where the annotations where made. Only intens inside the annotations_rectangle area are guaranteed to be annotated.

Note that, if a spot has a car, then the car is annotated on Cars JSON with the same car_id as the Spots JSON.

The paths described in the JSON follow the POSIX standart (e.g., root_directory/filename.extension)

Cars JSON

This type has the segmentations and car_id of each car. If an image doesn't have any car, then it's not in the JSON. Note that the same car on different images receive only one car_id.

Note that the following image has a reduced annotation_rectangle.

The file has a key "categories" which refers to the objects of the scope of the dataset. It also have the key "climates", which refers to the climate of each image. Then, the JSON have the key "images", that describes the informations of each image of a subset. Each of these informations of a image contains a unique key "id". The last key "annotations" refers to the annotation of each object, each of them also have a unique key "id". You can associate an object to an image by the "id" of an image and the "image_id" of an annotation.

The JSON structure is the following example

{
  "categories": [
    {
      "supercategory": "vehicle",
      "name": "car",
      "id": 0,
    }
  ],
  "climates": [
    {
      "id": 1,
      "name": "overcast",
    },
    {
      "id": 2,
      "name": "sunny",
    },
    {
      "id": 3,
      "name": "rainy",
    },
    {
      "id": 4,
      "name": "snowy",
    },
    {
      "id": 5,
      "name": "no_rain",
    },
    {
      "id": 6,
      "name": "undefined",
    }
  ],
  "images": [
    { 
      "id": 0,
      "file_name": "UFPR04/Sunny/2012-12-07/2012-12-07_20_17_28.jpg",
      "width": 1280,
      "height": 720,
      "annotationsRectangle": [upper_left_x, upper_left_y, width, height],
      "date": [2012, 12, 7],
      "time": [20, 17, 28],
      "climate": 2,
      "dataset": pklot,
      "subset": ufpr04,
    },
    {
      "id": 1,
      "file_name": "UFPR04/Rainy/2012-12-07/2012-12-07_16_57_25.jpg",
      ...
    },
    ...
  ],
  "annotations": [
    {
      "id": 0,
      "segmentation": [[x1, y1, x2, y2, ..., xn, yn], ..., [x1, y1, x2, y2, ..., xn, yn]],
      "area": area_in_pixels,
      "bbox": [upper_left_x, upper_left_y, width, height],
      "image_id": 1,
      "car_id": 500,
      "category_id": 0,
    },
    {
      "id": 1,
      "segmentation": [[x1, y1, x2, y2, ..., xn, yn], ..., [x1, y1, x2, y2, ..., xn, yn]],
      ...
    },
    ...
  ]
}

Spot JSON

This type has the segmentation of each parking lot, its occupation and corresponding car_id. Note that the car_id is the same of the corresponding car in Cars JSON.

The format is similar to the Cars JSON. Note that in each annotation of a spot there is a key "category_id", which is 0 if the spot is empty and 1 if it is occupied. If there's no car in the spot, the car_id is equal to -1. Also, if a spot is occupied with anything but a car, the car_id is equal to -2.

In order to cut a spot, there is a key "best_rotated_rect" on each annotation.

There is an example:

{
  "categories": [
    {
      "id": 0,
      "name": "empty",
    },
    {
      "id": 1,
      "name": "occupied",
    }
  ],
  "climates": [
    {
      "id": 1,
      "name": "overcast",
    },
    {
      "id": 2,
      "name": "sunny",
    },
    {
      "id": 3,
      "name": "rainy",
    },
    {
      "id": 4,
      "name": "snowy",
    },
    {
      "id": 5,
      "name": "no_rain",
    },
    {
      "id": 6,
      "name": "undefined",
    }
  ],
  "images": [
    { 
      "id": 0,
      "file_name": "UFPR04/Sunny/2012-12-07/2012-12-07_20_17_28.jpg",
      "width": 1280,
      "height": 720,
      "annotationsRectangle": [upper_left_x, upper_left_y, width, height],
      "date": [2012, 12, 7],
      "time": [20, 17, 28],
      "climate": 2,
      "dataset": pklot,
      "subset": ufpr04,
    },
    {
      "id": 1,
      "file_name": "UFPR04/Rainy/2012-12-07/2012-12-07_16_57_25.jpg",
      ...
    },
    ...
  ],
  "annotations": [
    {
      "id": 0,
      "segmentation": [[x1, y1, x2, y2, ..., xn, yn]],
      "bbox": [upper_left_x, upper_left_y, width, height],
      "image_id": 1,
      "car_id": 500,
      "category_id": 1,
      "best_rotated_rect": [[center_x, center_y], [width, height], angle]
    },
    {
      "id": 1,
      "segmentation": [[x1, y1, x2, y2, ..., xn, yn]],
      ...
    },
    ...
  ]
}

Changes from the PKLot to the MetaPKLot dataset

Corrections: Corrected the directory UFPR05/Sunny/2013-14-16 to UFPR05/Sunny/2013-04-16 Corrections: Day 2012-09-16 of PUCPR on climate Cloudy renamed to 2012-10-16

In removed_images/pklot.txt you can find find corrupted images that were removed from MetaPKLot dataset. In removed_images/cnr.txt you can find images with poor lighting, which were also removed. As well, the following images of PLds were removed due to repeated time or unexpected zoom.