2018年5月22日星期二

keras-yolo3推导保存图片和json

#! /usr/bin/env python
# -*- coding: utf-8 -*-
"""
Run a YOLO_v3 style detection model on test images.
"""
import cv2
import json
import colorsys
import os
import random
from timeit import time
from timeit import default_timer as timer  ### to calculate FPS

import numpy as np
from keras import backend as K
from keras.models import load_model
from PIL import Image, ImageFont, ImageDraw

from yolo3.model import yolo_eval
from yolo3.utils import letterbox_image

class YOLO(object):
    def __init__(self):
        self.model_path = 'model_data/yolo.h5'
        self.anchors_path = 'model_data/yolo_anchors.txt'
#        self.classes_path = 'model_data/coco_classes.txt'
        self.classes_path = 'model_data/voc_6c_classes.txt'
        self.score = 0.3
        self.iou = 0.5
        self.class_names = self._get_class()
        self.anchors = self._get_anchors()
        self.sess = K.get_session()
        self.model_image_size = (416, 416) # fixed size or (None, None)
        self.is_fixed_size = self.model_image_size != (None, None)
        self.boxes, self.scores, self.classes = self.generate()

    def _get_class(self):
        classes_path = os.path.expanduser(self.classes_path)
        with open(classes_path) as f:
            class_names = f.readlines()
        class_names = [c.strip() for c in class_names]
        return class_names

    def _get_anchors(self):
        anchors_path = os.path.expanduser(self.anchors_path)
        with open(anchors_path) as f:
            anchors = f.readline()
            anchors = [float(x) for x in anchors.split(',')]
            anchors = np.array(anchors).reshape(-1, 2)
        return anchors

    def generate(self):
        model_path = os.path.expanduser(self.model_path)
        assert model_path.endswith('.h5'), 'Keras model must be a .h5 file.'

        self.yolo_model = load_model(model_path, compile=False)
        print('{} model, anchors, and classes loaded.'.format(model_path))

        # Generate colors for drawing bounding boxes.
        hsv_tuples = [(x / len(self.class_names), 1., 1.)
                      for x in range(len(self.class_names))]
        self.colors = list(map(lambda x: colorsys.hsv_to_rgb(*x), hsv_tuples))
        self.colors = list(
            map(lambda x: (int(x[0] * 255), int(x[1] * 255), int(x[2] * 255)),
                self.colors))
        random.seed(10101)  # Fixed seed for consistent colors across runs.
        random.shuffle(self.colors)  # Shuffle colors to decorrelate adjacent classes.
        random.seed(None)  # Reset seed to default.

        # Generate output tensor targets for filtered bounding boxes.
        self.input_image_shape = K.placeholder(shape=(2, ))
        boxes, scores, classes = yolo_eval(self.yolo_model.output, self.anchors,
                len(self.class_names), self.input_image_shape,
                score_threshold=self.score, iou_threshold=self.iou)
        return boxes, scores, classes

    def detect_image(self, pathall):
        image = Image.open(pathall)
        start = time.time()
#        imgcv = cv2.imread(image)##########

        if self.is_fixed_size:
            assert self.model_image_size[0]%32 == 0, 'Multiples of 32 required'
            assert self.model_image_size[1]%32 == 0, 'Multiples of 32 required'
            boxed_image = letterbox_image(image, tuple(reversed(self.model_image_size)))
        else:
            new_image_size = (image.width - (image.width % 32),
                              image.height - (image.height % 32))
            boxed_image = letterbox_image(image, new_image_size)
        image_data = np.array(boxed_image, dtype='float32')

        print(image_data.shape)
        image_data /= 255.
        image_data = np.expand_dims(image_data, 0)  # Add batch dimension.

        out_boxes, out_scores, out_classes = self.sess.run(
            [self.boxes, self.scores, self.classes],
            feed_dict={
                self.yolo_model.input: image_data,
                self.input_image_shape: [image.size[1], image.size[0]],
                K.learning_phase(): 0
            })

        print('Found {} boxes for {}'.format(len(out_boxes), 'img'))

        font = ImageFont.truetype(font='font/FiraMono-Medium.otf',
                    size=np.floor(3e-2 * image.size[1] + 0.5).astype('int32'))
        thickness = (image.size[0] + image.size[1]) // 300
       
        resultsForJSON = []#+++

        for i, c in reversed(list(enumerate(out_classes))):
            predicted_class = self.class_names[c]
            box = out_boxes[i]
            score = out_scores[i]

            label = '{} {:.2f}'.format(predicted_class, score)
            draw = ImageDraw.Draw(image)
            label_size = draw.textsize(label, font)

            top, left, bottom, right = box
            top = max(0, np.floor(top + 0.5).astype('int32'))
            left = max(0, np.floor(left + 0.5).astype('int32'))
            bottom = min(image.size[1], np.floor(bottom + 0.5).astype('int32'))
            right = min(image.size[0], np.floor(right + 0.5).astype('int32'))
            print(label, (left, top), (right, bottom))
           
            area=(bottom-top)*(right-left)#####+
            resultsForJSON.append({"label": predicted_class,\
                                   "confidence": float('%.2f' % score),\
                                   "topleft": {"x": float(left), "y": float(top)},\
                                   "bottomright": {"x": float(right), "y": float(bottom)},\
                                   "area":float(area)})
   
            if top - label_size[1] >= 0:
                text_origin = np.array([left, top - label_size[1]])
            else:
                text_origin = np.array([left, top + 1])

            # My kingdom for a good redistributable image drawing library.
            for i in range(thickness):
                draw.rectangle(
                    [left + i, top + i, right - i, bottom - i],
                    outline=self.colors[c])
            draw.rectangle(
                [tuple(text_origin), tuple(text_origin + label_size)],
                fill=self.colors[c])
            draw.text(text_origin, label, fill=(0, 0, 0), font=font)
            del draw

        end = time.time()
        print(end - start)
        ####################保存json##########start
        print(resultsForJSON)
        imgdir = "/Users/sisyphus/keras-yolo3/"    ###########+
        outfolder = os.path.join(imgdir, 'out')
        img_name = os.path.join(outfolder, os.path.basename(pathall))
#        cv2.imwrite(img_name, image)
        if resultsForJSON == []:
            resultsForJSON.append({"label": "Normal",\
                                   "confidence": float(0), \
                                   "topleft": {"x": float(0), "y": float(0)}, \
                                   "bottomright": {"x": float(0), "y": float(0)},\
                                   "area":float(0)})
            textJSON = json.dumps(resultsForJSON)
            textFile = os.path.splitext(img_name)[0] + ".json"
            print('\n')
            with open(textFile, 'w') as f:
                f.write(textJSON)
            print('Normal\n')
        else:
            textJSON = json.dumps(resultsForJSON)
            textFile = os.path.splitext(img_name)[0] + ".json"
            print(textFile)
            print('\n')
            with open(textFile, 'w') as f:
                f.write(textJSON)
        ####################保存json##########end       
        return image

    def close_session(self):
        self.sess.close()
       
       
    def detect_image_video(self, image, curr_fps):
        start = time.time()

        if self.is_fixed_size:
            assert self.model_image_size[0]%32 == 0, 'Multiples of 32 required'
            assert self.model_image_size[1]%32 == 0, 'Multiples of 32 required'
            boxed_image = letterbox_image(image, tuple(reversed(self.model_image_size)))
        else:
            new_image_size = (image.width - (image.width % 32),
                              image.height - (image.height % 32))
            boxed_image = letterbox_image(image, new_image_size)
        image_data = np.array(boxed_image, dtype='float32')

        print(image_data.shape)
        image_data /= 255.
        image_data = np.expand_dims(image_data, 0)  # Add batch dimension.

        out_boxes, out_scores, out_classes = self.sess.run(
            [self.boxes, self.scores, self.classes],
            feed_dict={
                self.yolo_model.input: image_data,
                self.input_image_shape: [image.size[1], image.size[0]],
                K.learning_phase(): 0
            })

        print('Found {} boxes for {}'.format(len(out_boxes), 'img'))

        font = ImageFont.truetype(font='font/FiraMono-Medium.otf',
                    size=np.floor(3e-2 * image.size[1] + 0.5).astype('int32'))
        thickness = (image.size[0] + image.size[1]) // 300
       
        resultsForJSON = []#+++

        for i, c in reversed(list(enumerate(out_classes))):
            predicted_class = self.class_names[c]
            box = out_boxes[i]
            score = out_scores[i]

            label = '{} {:.2f}'.format(predicted_class, score)
            draw = ImageDraw.Draw(image)
            label_size = draw.textsize(label, font)

            top, left, bottom, right = box
            top = max(0, np.floor(top + 0.5).astype('int32'))
            left = max(0, np.floor(left + 0.5).astype('int32'))
            bottom = min(image.size[1], np.floor(bottom + 0.5).astype('int32'))
            right = min(image.size[0], np.floor(right + 0.5).astype('int32'))
            print(label, (left, top), (right, bottom))
           
            area=(bottom-top)*(right-left)#####+
            resultsForJSON.append({"label": predicted_class,\
                                   "confidence": float('%.2f' % score),\
                                   "topleft": {"x": float(left), "y": float(top)},\
                                   "bottomright": {"x": float(right), "y": float(bottom)},\
                                   "area":float(area)})

            if top - label_size[1] >= 0:
                text_origin = np.array([left, top - label_size[1]])
            else:
                text_origin = np.array([left, top + 1])

            # My kingdom for a good redistributable image drawing library.
            for i in range(thickness):
                draw.rectangle(
                    [left + i, top + i, right - i, bottom - i],
                    outline=self.colors[c])
            draw.rectangle(
                [tuple(text_origin), tuple(text_origin + label_size)],
                fill=self.colors[c])
            draw.text(text_origin, label, fill=(0, 0, 0), font=font)
            del draw

        end = time.time()
        print(end - start)
       
        ####################保存json##########start
        print(resultsForJSON)
        imgdir = "/Users/sisyphus/keras-yolo3/"    ###########+
        outfolder = os.path.join(imgdir, 'video_out/')
#        img_name = os.path.join(outfolder, curr_fps)
#        cv2.imwrite(img_name, image)
        if resultsForJSON == []:
            resultsForJSON.append({"label": "Normal",\
                                   "confidence": float(0), \
                                   "topleft": {"x": float(0), "y": float(0)}, \
                                   "bottomright": {"x": float(0), "y": float(0)},\
                                   "area":float(0)})
            textJSON = json.dumps(resultsForJSON)
            textFile = outfolder + str(curr_fps) + ".json"
            print('\n')
            with open(textFile, 'w') as f:
                f.write(textJSON)
            print('Normal\n')
        else:
            textJSON = json.dumps(resultsForJSON)
            textFile = outfolder + str(curr_fps) + ".json"
            print(textFile)
            print('\n')
            with open(textFile, 'w') as f:
                f.write(textJSON)
        ####################保存json##########end
        ####################保存图片##########start
        imgdir = "/Users/sisyphus/keras-yolo3/"    ###########+
        outfolder = os.path.join(imgdir, 'video_out/')
        img_name = outfolder + str(curr_fps) + ".jpg"
        image.save(img_name)
        ####################保存图片##########end
        return image

    def close_session(self):
        self.sess.close()
       
       


def detect_video(yolo, video_path):
    import cv2
    vid = cv2.VideoCapture(video_path)
    if not vid.isOpened():
        raise IOError("Couldn't open webcam or video")
    accum_time = 0
    curr_fps = 0
    fps = "FPS: ??"
    prev_time = timer()
    while True:
        return_value, frame = vid.read()
#        frame = np.asarray(frame)####+
        image = Image.fromarray(frame)
        image = yolo.detect_image_video(image, curr_fps)
        result = np.asarray(image)
        curr_time = timer()
        exec_time = curr_time - prev_time
        prev_time = curr_time
        accum_time = accum_time + exec_time
        curr_fps = curr_fps + 1
        print(str(curr_fps)+"########")
        if accum_time > 1:
            accum_time = accum_time - 1
            fps = "FPS: " + str(curr_fps)
#            curr_fps = 0
        cv2.putText(result, text=fps, org=(3, 15), fontFace=cv2.FONT_HERSHEY_SIMPLEX,
                    fontScale=0.50, color=(255, 0, 0), thickness=2)
        cv2.namedWindow("result", cv2.WINDOW_NORMAL)
        cv2.imshow("result", result)
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break
#        choice = cv2.waitKey(1)
#        if choice == 27: break
#    cap.release()
#    cv2.destroyAllWindows()
    yolo.close_session()
   



#def detect_img(yolo):
#    while True:
#        img = input('Input image filename:')
##        img = "/Users/sisyphus/keras-yolo3/sample_img_test30/000271.jpg"
#        try:
#            image = Image.open(img)
#        except:
#            print('Open Error! Try again!')
#            continue
#        else:
#            r_image = yolo.detect_image(image)
#            r_image.show()
#    yolo.close_session()
   
#####################################单张图片预测   
#def detect_img(yolo):
#    img = "/Users/sisyphus/keras-yolo3/sample_img_test30/000271.jpg"
#    image = Image.open(img)
#    r_image = yolo.detect_image(image)
#    r_image.show()
   
   
####################################文件夹内所有图片预测
def detect_img(yolo):
    imgdir = "/Users/sisyphus/keras-yolo3/"    ###########+
    outfolder = os.path.join(imgdir, 'out')
    path = "/Users/sisyphus/keras-yolo3/sample_img_test1"
    for xd in os.listdir(path):
        print(xd)
        print(os.path.join(path,xd))
        pathall = os.path.join(path,xd)
        r_image = yolo.detect_image(pathall)
#        r_image.show()
        img_name = os.path.join(outfolder, os.path.basename(pathall))
        r_image.save(img_name)
           



if __name__ == '__main__':
    detect_img(YOLO())

没有评论:

发表评论

Failed to find TIFF library

ImportError: Failed to find TIFF library. Make sure that libtiff is installed and its location is listed in PATH|LD_LIBRARY_PATH|.. 解决方法: ...