Skip to content

Latest commit

 

History

History
108 lines (71 loc) · 3.82 KB

readme.md

File metadata and controls

108 lines (71 loc) · 3.82 KB
Raw

Knowledge Graph Input

This project defines the data format of the video_labels, which are accepted as a input data for the Knowledge Graph Builder project.

Since Knowledge Graph Builder takes JSONLines file that is composed of video_label objects only, please validate your input data file before using with Knowledge Graph Builder.

Definition of the video_label object

The main purpose of this project is to define the data format of video_label object. This section contains both formal and informal definition of this object.

Description

A video_label object should follow one of the following 6 types of the objects.

Video label objects

Object

  • {"type": "object", "class": "person", "label": "Person A", "seconds": 15.0, "coordinates": [100, 200, 20, 30]}
  • {"type": "object", "id": "person_ross_geller", "class": "person", "label": "Ross Geller", "seconds": 15.0, "coordinates": [100, 200, 20, 30]}

Behavior

  • {"type": "behavior", "class": "stand", "seconds": 15.0, "object": ObjectIndicator}

Emotion

  • {"type": "emotion", "class": "happy", "seconds": 15.0, "object": ObjectIndicator}

Relation

  • {"type": "relation", "class": RelationalClass, "subclass": "wear", "seconds": 15.0, "source": ObjectIndicator, "target": ObjectIndicator}

Location

  • {"type": "location", "class": "central_perk", "seconds": 15.0}

Sound

  • {"type": "sound", "class": "glass_crashing", "seconds": 15.0}

Sub-objects in video label objects

ObjectIndicator

  • {"id": "person_ross_geller"} or
  • {"coordinates": [100, 200, 20, 30]}

RelationalClass

  • "behavior" for subclass "wear", "hold", "hug", …
  • "emotion" for subclass "love", "like", …
  • "position" for subclass "above", "below", "next_to", …
  • "social" for subclass "son_of", "father_of", "lover_of", "friend_of", …

Formal definition

A formal definition of the video_label object can be found in video_label.schema.json file. This file defines the format of the video_label object in JSON Schema.

Validation of the input data file

This project includes a basic validator script to check if a given jsonlines file is a valid input data for Knowledge Graph Builder. Instructions for using the validator is as follows.

Before using the validator, make sure that Python 3 is available in your local environment.

Set up a virtual environment for Python 3.

knowledge-graph-input$ python3 -m venv ./env

Activate the virtual environment to use.

knowledge-graph-input$ source ./env/bin/activate

Install required dependencies with Pip in your virtual environment.

(env) knowledge-graph-input$ pip install -r requirements.txt

Run validate.py script with your input data file as standard input. Example command will validate the test input data file test_input.jsonl.

(env) knowledge-graph-input$ python3 validate.py < test_input.jsonl
Done.

If there are errors in the input data, erronous lines are shown on the terminal as standard error output.

(env) knowledge-graph-input jeongmin$ python3 validate.py < test_input.jsonl
Validation error at line 3:
"{"type": "locational", "class": "central_perk", "seconds": 0.5}"
Validation error at line 14:
"{"type": "behavior", "seconds": 2.5, "object": {"id": "person_ross_geller"}}"
Done.

Acknowledgements

This work was supported by Institute for Information & communications Technology Promotion(IITP) grant funded by the Korea government(MSIT) (2017-0-01780, The technology development for event recognition/relational reasoning and learning knowledge based system for video understanding)