##Evaluation
Main files: evaluate.py evaluate_summary.py
evaluate.py is a Python 2.7 script intended to compare exactly one pair of annotation files.
Requires a ground truth annotation file, a non-ground truth annotation file, a config file, and an output path.
Optional flag: --json
Tells the script to use .json files in place of .ann files as arguments. NOTE: JSON object structure not finalized; functionality incomplete.
Run "python evaluate.py --help" for more information on the individual flags.
evaluate_summary.py runs the evaluation script on all pairs of ground truth and non-ground truth files within a folder that follow a specific naming convention.
Requires a directory containing all annotation files and a directory to contain outputs.
The naming convention is as follows:
- Ground truth annotation files must be named "gt*.ann", where the asterisk represents an arbitrary sequence of characters.
- Non-ground truth annotation files must be named "res*.ann", where the asterisk represents an arbitrary sequence of characters.
NOTE: For evaluation to be run on a pair of files, the arbitrary sequences of characters used in each filename must be identical. Otherwise, the files will be discarded.