psenet的eval_ctw1500.py解析

评价模型指标，自带的eval_ctw1500.py是根据两两框重叠面积大于0.5算正例tp。其中用到了import Polygon as plg 模块，可以方便的处理多边形的重叠计算面积
1.pip install Polygon2
2.cover = set() 这里用到了set，之前在c++里面也看到过这个的，一查发现差不多功能，就是没有重复的无序的

file_util.py

import os

def read_dir(root):
	file_path_list = []
	for file_path, dirs, files in os.walk(root):
		for file in files:
			file_path_list.append(os.path.join(file_path, file).replace('\', '/'))
	file_path_list.sort()
	return file_path_list

def read_file(file_path):
	file_object = open(file_path, 'r')
	file_content = file_object.read()
	file_object.close()
	return file_content

def write_file(file_path, file_content):
	if file_path.find('/') != -1:
		father_dir = '/'.join(file_path.split('/')[0:-1])
		if not os.path.exists(father_dir):
			os.makedirs(father_dir)
	file_object = open(file_path, 'w')
	file_object.write(file_content)
	file_object.close()


def write_file_not_cover(file_path, file_content):
	father_dir = '/'.join(file_path.split('/')[0:-1])
	if not os.path.exists(father_dir):
		os.makedirs(father_dir)
	file_object = open(file_path, 'a')
	file_object.write(file_content)
	file_object.close()

eval_ctw1500.py

import file_util
import Polygon as plg
import numpy as np

pred_root = '../../outputs/submit_ctw1500/'
gt_root = '../../data/CTW1500/test/text_label_curve/'

def get_pred(path):
    lines = file_util.read_file(path).split('
')
    bboxes = []
    for line in lines:
        if line == '':
            continue
        bbox = line.split(',')
        if len(bbox) % 2 == 1:
            print path
        bbox = [(int)(x) for x in bbox]
        bboxes.append(bbox)
    return bboxes

def get_gt(path):
    print("###############################path=%s"%(path))
    lines = file_util.read_file(path).split('
')
    bboxes = []
    for line in lines:
        if line == '':
            continue
        # line = util.str.remove_all(line, 'xefxbbxbf')
        # gt = util.str.split(line, ',')
        gt = line.split(',')

        x1 = np.int(gt[0])
        y1 = np.int(gt[1])

        bbox = [np.int(gt[i]) for i in range(4, 32)]
        bbox = np.asarray(bbox) + ([x1, y1] * 14)
        
        bboxes.append(bbox)
    return bboxes

def get_union(pD,pG):
    areaA = pD.area();
    areaB = pG.area();
    return areaA + areaB - get_intersection(pD, pG);        

def get_intersection(pD,pG):
    pInt = pD & pG
    if len(pInt) == 0:
        return 0
    return pInt.area()

if __name__ == '__main__':
    th = 0.5
    pred_list = file_util.read_dir(pred_root)

    tp, fp, npos = 0, 0, 0
    
    for pred_path in pred_list:
        preds = get_pred(pred_path)
        gt_path = gt_root + pred_path.split('/')[-1]
        gts = get_gt(gt_path)
        npos += len(gts)
        
        cover = set()
        for pred_id, pred in enumerate(preds):
            pred = np.array(pred)
            pred = pred.reshape(pred.shape[0] / 2, 2)
            # if pred.shape[0] <= 2:
            #     continue
            pred_p = plg.Polygon(pred)
            
            flag = False
            for gt_id, gt in enumerate(gts):
                gt = np.array(gt)
                gt = gt.reshape(gt.shape[0] / 2, 2)
                gt_p = plg.Polygon(gt)

                union = get_union(pred_p, gt_p)
                inter = get_intersection(pred_p, gt_p)

                if inter * 1.0 / union >= th:
                    if gt_id not in cover:
                        flag = True
                        cover.add(gt_id)
            if flag:
                tp += 1.0
            else:
                fp += 1.0

    print tp, fp, npos
    precision = tp / (tp + fp)
    recall = tp / npos
    hmean = 0 if (precision + recall) == 0 else 2.0 * precision * recall / (precision + recall)

    print 'p: %.4f, r: %.4f, f: %.4f'%(precision, recall, hmean)