utils.py

import numpy as np
from nms import nms
import cfg
from shapely.geometry import Polygon


class Averager(object):
    """Compute average for torch.Tensor, used for loss average."""

    def __init__(self):
        self.reset()

    def add(self, v):
        count = v.data.numel()
        v = v.data.sum()
        self.n_count += count
        self.sum += v

    def reset(self):
        self.n_count = 0
        self.sum = 0

    def val(self):
        res = 0
        if self.n_count != 0:
            res = self.sum / float(self.n_count)
        return res


class eval_pre_rec_f1(object):
    '''输入每个batch的预测结果跟图片真实矩形框，计算查准率precision/召回率recall/F1 score'''

    def __init__(self):
        self.pixel_threshold = float(cfg.pixel_threshold)
        self.reset()

    def reset(self):
        self.img_num = 0
        self.pre = 0
        self.rec = 0
        self.f1_score = 0

    def val(self):
        mpre = self.pre / self.img_num * 100
        mrec = self.rec / self.img_num * 100
        mf1_score = self.f1_score / self.img_num * 100
        return mpre, mrec, mf1_score

    def sigmoid(self, x):
        """`y = 1 / (1 + exp(-x))`"""
        return 1 / (1 + np.exp(-x))

    def get_iou(self, g, p):
        g = Polygon(g)
        p = Polygon(p)
        if not g.is_valid or not p.is_valid:
            return 0
        inter = Polygon(g).intersection(Polygon(p)).area
        union = g.area + p.area - inter
        if union == 0:
            return 0
        else:
            return inter/union

    def eval_one(self, quad_scores, quad_after_nms, gt_xy, quiet=cfg.quiet):
        num_gts = len(gt_xy)
        quad_scores_no_zero = []  # 剔除残缺quad，并储存每个quad的score
        quad_after_nms_no_zero = []  # 剔除残缺quad
        for score, geo in zip(quad_scores, quad_after_nms):
            if np.amin(score) > 0:
                quad_scores_no_zero.append(sum(score))
                quad_after_nms_no_zero.append(geo)
            elif not quiet:
                print('quad invalid with vertex num less then 4.')
                continue
        num_quads = len(quad_after_nms_no_zero)
        if num_quads == 0:
            return 0, 0, 0
        quad_flag = np.zeros(num_quads)  # 记录quad是否被匹配
        gt_flag = np.zeros(num_gts)  # 记录gt是否被匹配
        # print(num_quads, '-------', num_gts)
        quad_scores_no_zero = np.array(quad_scores_no_zero)
        scores_idx = np.argsort(quad_scores_no_zero)[::-1]  # 记录quad_scores从大到小坐标
        for i in range(num_quads):
            idx = scores_idx[i]
            #  score = quad_scores_no_zero[idx]
            geo = quad_after_nms_no_zero[idx]  # 按score值从大到小依次取出对应矩形框
            for j in range(num_gts):
                if gt_flag[j] == 0:
                    gt_geo = gt_xy[j]
                    iou = self.get_iou(geo, gt_geo)
                    if iou >= cfg.iou_threshold:
                        gt_flag[j] = 1  # 记录被匹配的gt框
                        quad_flag[i] = 1  # 记录被匹配的quad框
        tp = np.sum(quad_flag)
        fp = num_quads - tp
        fn = num_gts - tp
        pre = tp / (tp + fp)  # 查准率
        rec = tp / (tp + fn)  # 查全率
        if pre + rec == 0:
            f1_score = 0
        else:
            f1_score = 2 * pre * rec / (pre + rec)
        # print(pre, '---', rec, '---', f1_score)
        return pre, rec, f1_score

    def add(self, out, gt_xy_list):
        self.img_num += len(gt_xy_list)
        ys = out.cpu().detach().numpy()  # (N, 7, 64, 64)
        if ys.shape[1] == 7:
            ys = ys.transpose((0, 2, 3, 1))  # NCHW->NHWC
        for y, gt_xy in zip(ys, gt_xy_list):  # 取出每张图片的预测结果与矩形框
            y[:, :, :3] = self.sigmoid(y[:, :, :3])
            cond = np.greater_equal(y[:, :, 0], self.pixel_threshold)
            activation_pixels = np.where(cond)
            quad_scores, quad_after_nms = nms(y, activation_pixels)
            # nms返回的quad_scores为：[[a, a, b, b], [c, c, d, d]...]
            # 每个矩形框返回四个score，四个score中头两个相同，后两个相同分别代表头部跟尾部的分数
            if (len(quad_after_nms) == 0) or (sum(sum(quad_scores)) == 0):
                if not cfg.quiet:
                    print('NMS后不存在矩形框！！')
                continue
            else:
                pre, rec, f1_score = self.eval_one(quad_scores, quad_after_nms, gt_xy)
                self.pre += pre
                self.rec += rec
                self.f1_score += f1_score