evaluate_model2.py

import utils.generator
import utils.unet
import numpy as np
import pandas as pd
import sklearn.metrics

def perf_measure(gt, pred):
    pred = np.round(pred).astype(int).flatten()
    gt = gt.flatten()
    TP = 0
    FP = 0
    TN = 0
    FN = 0

    for i in range(len(pred)): 
        if gt[i]==pred[i]==1:
           TP += 1
        if pred[i]==1 and gt[i]!=pred[i]:
           FP += 1
        if gt[i]==pred[i]==0:
           TN += 1
        if pred[i]==0 and gt[i]!=pred[i]:
           FN += 1

    return(TP, FP, TN, FN)

def evaluate(pred, gt):
    pred = np.round(pred).astype(int).flatten()
    gt = gt.flatten()

    fmes = sklearn.metrics.f1_score(gt, pred)
    acc = sklearn.metrics.accuracy_score(gt, pred)
    rec = sklearn.metrics.recall_score(gt, pred)
    jacc = sklearn.metrics.jaccard_score(gt, pred)
    TP, FP, TN, FN = perf_measure(gt, pred)
    return fmes, acc, rec, jacc, TP, FP, TN, FN 


def main(img_path, gt_path, selected_imgs, model_path, out_path, wo_crf, depth):
    valid_gen = utils.generator.DataGenerator(img_path, gt_path,
                                              include=selected_imgs,
                                              augment=False)
    if wo_crf:
        unet = utils.unet.XceptionUNet(valid_gen.input_shape, depth=depth)
        unet.model.load_weights(model_path)
        model = unet.model
    else:
        unet = utils.unet.XceptionUNetCRF(valid_gen.input_shape, iterations=20)
        unet.crf_model.load_weights(model_path)
        model = unet.crf_model

    results = {'fmes': [], 'acc': [], 'rec': [], 'jacc': [], 'TP': [], 'FP': [], 'TN': [], 'FN': []}
    valid_it = iter(valid_gen)

    for img, gt in valid_it:
        out = model.predict(img)
        out = out[0, :, 1]  # .reshape(shape)
        gt = gt[0, :, 1]
        fmes, acc, rec, jacc, TP, FP, TN, FN = evaluate(out, gt)
        results['fmes'].append(fmes)
        results['acc'].append(acc)
        results['rec'].append(rec)
        results['jacc'].append(jacc)
        results['TP'].append(TP)
        results['FP'].append(FP)
        results['TN'].append(TN)
        results['FN'].append(FN)

    df = pd.DataFrame(results)
    df.to_csv(out_path, index=False)


if __name__ == '__main__':
    import argparse

    parser = argparse.ArgumentParser(
                       description='Evaluate model on given images',
                       formatter_class=argparse.ArgumentDefaultsHelpFormatter)
    parser.add_argument('img_path', help='Path to image folder')
    parser.add_argument('gt_path')
    parser.add_argument('selected_imgs', help='Path to CSV file with '
                        'selected image indices')
    parser.add_argument('model_path')
    parser.add_argument('out_path', help='Path to output CSV file')
    parser.add_argument('--wo_crf', action='store_true')
    parser.add_argument('--depth', type=int, default=2)

    args = vars(parser.parse_args())
    main(**args)