get_dr

from keras.layers import Input
from efficientdet import EfficientDet
from PIL import Image
from utils.utils import BBoxUtility,letterbox_image,efficientdet_correct_boxes
from utils.anchors import get_anchors
import math
import copy
import json
import numpy as np
import os
def preprocess_input(image):
    image /= 255
    mean = [0.485, 0.456, 0.406]
    std = [0.229, 0.224, 0.225]
    image -= mean
    image /= std
    return image
class mAP_EfficientDet(EfficientDet):
    #---------------------------------------------------#
    #   检测图片
    #---------------------------------------------------#
    def detect_image(self,image_id,image):

        ##coco的数据box是下x,y,w,h

        self.confidence = 0.05  #起始的置信度
        f = open("./input/detection-results/"+image_id+".txt","w") 
        image_shape = np.array(np.shape(image)[0:2])

        crop_img = letterbox_image(image, [self.model_image_size[0],self.model_image_size[1]])
        photo = np.array(crop_img,dtype = np.float32)

        # 图片预处理，归一化
        photo = np.reshape(preprocess_input(photo),[1,self.model_image_size[0],self.model_image_size[1],self.model_image_size[2]])

        preds = self.Efficientdet.predict(photo)

        # 将预测结果进行解码
        results = self.bbox_util.detection_out(preds, self.prior, confidence_threshold=self.confidence)
        if len(results[0])<=0:
            return image
        results = np.array(results)
        
        # 筛选出其中得分高于confidence的框
        det_label = results[0][:, 5]
        det_conf = results[0][:, 4]
        det_xmin, det_ymin, det_xmax, det_ymax = results[0][:, 0], results[0][:, 1], results[0][:, 2], results[0][:, 3]
        
        top_indices = [i for i, conf in enumerate(det_conf) if conf >= self.confidence]
        top_conf = det_conf[top_indices]
        top_label_indices = det_label[top_indices].tolist()
        top_xmin, top_ymin, top_xmax, top_ymax = np.expand_dims(det_xmin[top_indices],-1),np.expand_dims(det_ymin[top_indices],-1),np.expand_dims(det_xmax[top_indices],-1),np.expand_dims(det_ymax[top_indices],-1)
        
        # 去掉灰条
        boxes = efficientdet_correct_boxes(top_ymin,top_xmin,top_ymax,top_xmax,np.array([self.model_image_size[0],self.model_image_size[1]]),image_shape)

        for i, c in enumerate(top_label_indices):
            predicted_class = self.class_names[int(c)]
            score =str(top_conf[i])

            top, left, bottom, right = boxes[i]
            image_result = {
                'image_id': image_id,  # 图片名称
                'category_id': predicted_class,  # 图片类别信息
                'score': float(score[:6]),
                'bbox':[(left+right)/2.0,(top+bottom)/2.0,right-left,bottom-top],
            }

            # f.write("%s %s %s %s %s %s\n" % (predicted_class, score[:6], str(int(left)), str(int(top)), str(int(right)),str(int(bottom))))
            results_json.append(image_result)
        # f.close()

        return 

efficientdet = mAP_EfficientDet()
image_ids = open('VOCdevkit/VOC2007/ImageSets/Main/test.txt').read().strip().split()

if not os.path.exists("./input"):
    os.makedirs("./input")
if not os.path.exists("./input/detection-results"):
    os.makedirs("./input/detection-results")
if not os.path.exists("./input/images-optional"):
    os.makedirs("./input/images-optional")

results_json=[]

imgIds= []
for image_id in image_ids:
    image_path = "./VOCdevkit/VOC2007/JPEGImages/"+image_id+".jpg"
    image = Image.open(image_path)
    image.save("./input/images-optional/"+image_id+".jpg")  #把对应的图片取出来
    efficientdet.detect_image(image_id,image)
    imgIds.append(image_id)
    print(image_id," done!")

# write output
json.dump(results_json, open('test_bbox_results.json', 'w'), indent=4)

print("Conversion completed!")
print(1)