evaluate_outputs.py

import argparse
import pandas as pd
import utils
import numpy as np

box_names = []

for i in range(1, 61):
    box_names.append('x1_' + str(i))
    box_names.append('y1_' + str(i))
    box_names.append('x2_' + str(i))
    box_names.append('y2_' + str(i))

parser = argparse.ArgumentParser()
parser.add_argument('-gt', help='Path to ground truth directory',default = './outputs/ground_truth/')
parser.add_argument('-pred', help='Path to prediction directory',default = './outputs/sted/')
args = parser.parse_args()

# for detector in ['yolo', 'mask-rcnn']:
for detector in ['yolo']:
    aious = []
    fious = []
    ades = []
    fdes = []
    for fold in [1, 2, 3]:

        gt_df = pd.read_csv(args.gt + '/test_' + detector + '_fold_' +
                            str(fold) + '.csv')
        pred_df = pd.read_csv(args.pred + '/test_' + detector + '_fold_' +
                              str(fold) + '.csv')

        gt_boxes = gt_df[box_names].values.reshape(len(gt_df), -1, 4)
        pred_boxes = pred_df[box_names].values.reshape(len(pred_df), -1, 4)

        ade = utils.calc_ade(gt_boxes, pred_boxes)
        fde = utils.calc_fde(gt_boxes, pred_boxes)
        aiou = utils.calc_aiou(gt_boxes, pred_boxes)
        fiou = utils.calc_fiou(gt_boxes, pred_boxes)

        aious.append(aiou)
        fious.append(fiou)
        ades.append(ade)
        fdes.append(fde)

        print()
        print(detector + ' fold ' + str(fold))
        print('AIOU: ', round(aiou * 100, 1))
        print('FIOU: ', round(fiou * 100, 1))
        print('ADE:  ', round(ade, 1))
        print('FDE:  ', round(fde, 1))

    print()
    print('Mean over 3 folds for ' + detector)
    print('AIOU: ', round(np.mean(aious) * 100, 1))
    print('FIOU: ', round(np.mean(fious) * 100, 1))
    print('ADE:  ', round(np.mean(ades), 1))
    print('FDE:  ', round(np.mean(fdes), 1))