Skip to content

Latest commit

 

History

History
50 lines (46 loc) · 1.83 KB

README.md

File metadata and controls

50 lines (46 loc) · 1.83 KB
Raw

Data Generate

CAD-60 Dataset

To genrate json file

Open the file CAD/generate_json-CAD60.ipynb in your jupyter-notebook. Befoer running the code, you have to change the default location which stores your dataset in code line 2.

data_dicts = r"./"

Dictionary structure

{
    "a str of ten-digit integer" : {
        "person_ID": int
        "folder_path": image folder,
        "action_label": string of action label,
        "frames": {
            "ID" : {
                "ROI" : {
                    "Joint ID" : {
                        "orin" : list,
                        "CONF" : int
                    }
                }
                "P" : {
                    "Joint ID" : {
                        "coor" : list,
                        "CONF" : int
                    }
                }
            }
        }
    }
}
  • "a str of ten-digit integer" is a string of code which represends a specific image folder. It is corresponding to one activity.
  • "person_ID" is an integer between 1,2,3,4 due to the data were collected from four person.
  • "folder_path" is the location of image folder. In CAD-60 dataset, RGB and depth images are sotred in one folder. The folder code is linked to a str of ten0digit interger.
  • "action_label" refers to the action that it belongs to.
  • "frames" stores all data in "folder_path", you can search by image ID.
  • "ID" is the interger begins from 1. It is refers to the image ID.
  • "ROI" is orientation of the joints.
  • "Joint ID" is an interger begins from 0, end in 10. In CAD dataset, there are only 11 joints include the orientation.
  • "orin" is a list with 9 elements.
  • "CONF" is the boolean confidence value (0 or 1).
  • "P" refers to joint points.
  • "coor" is a list with 3 elements, which are x,y,z. The origin of the corridinate system is at the sensor.