Annotation files given in BIWI dataset (http://www.vision.ee.ethz.ch/en/datasets/)
Info : Video frame rate: 2.5fps (6 frames = 0.04s)
Annotation files given in crowds dataset (https://graphics.cs.ucy.ac.cy/research/downloads/crowd-data)
Info: Video frame rate: 25fps (1 frame = 0.04s).
In this repository you will find the Python script and preprocessed data where:
Data_Formated_.csv:
Extracted data from (obsmat.txt for ETH data, .VSP file from the UCY dataset) and adjusted to the desired format which is [FrameID, PedID, PosX,PosY]
(meters if its from the ETH dataset, and pixels if its from the UCY dataset)
Data_in_meters_.csv:
Data converted from image coordinates to world coordinates.
(meters)
interpolated_data_.csv:
Interpolate data from Data_in_meters for each frame.
(meters)
Final_data_.csv:
Extract interpolated_data at every 10 frames (2.5fps)
(meters)
UCY Preprocessing.ipynb, ETH Preprocessing.ipynb:
Python scripts for the preprocessing steps for each dataset.
KNIME Analytics platform workflow:
A workflow for extracting data analytics and data cleaning.