Python 例如，tensorflow对象检测失败

---

系统信息

• 您正在使用的模型的顶级目录是什么：tensorflow/models
• 我是否编写了自定义代码（而不是使用TensorFlow中提供的股票示例脚本）：否
• 操作系统平台和发行版（如Linux Ubuntu 16.04）：Windows 10 x64
• TensorFlow安装自（源代码或二进制文件）：二进制文件
• TensorFlow版本（使用下面的命令）：1.3.0
• Bazel版本（如果从源代码处编译）：
• CUDA/cuDNN版本：
• GPU型号和内存：
• 复制的精确命令：python对象检测\eval.py--logtostderr--checkpoint\u dir=train--eval\u dir=eval--pipeline\u config\u path=ssd\u mobilenet\u v1.config & python.py

描述问题 使用COCO集合中的finetune检查点对Oxford IIIT Pet数据集进行训练时，我在运行eval.py脚本时收到以下错误消息：

损坏的JPEG数据：标记0xd9之前有245个额外字节

在评估失败后运行detect.py脚本时，我得到了附加的图像，没有任何检测框

源代码/日志 detect.py的代码：

import numpy as np
import os
import six.moves.urllib as urllib
import sys
import tensorflow as tf
import zipfile
from object_detection.utils import label_map_util
from object_detection.utils import visualization_utils as vis_util
from collections import defaultdict
from io import StringIO
from matplotlib import pyplot as plt
from PIL import Image

PATH_TO_CKPT = os.path.join('inference_graphs', 'frozen_inference_graph.pb')
PATH_TO_LABELS = 'pet_label_map.pbtxt'
PATH_TO_TEST_IMAGES_DIR = os.path.join('test')
NUM_CLASSES = 37

detection_graph = tf.Graph()
with detection_graph.as_default():
  od_graph_def = tf.GraphDef()
  with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
    serialized_graph = fid.read()
    od_graph_def.ParseFromString(serialized_graph)
    tf.import_graph_def(od_graph_def, name='')

label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES, use_display_name=True)
category_index = label_map_util.create_category_index(categories)

def load_image_into_numpy_array(image):
  (im_width, im_height) = image.size
  return np.array(image.getdata()).reshape(
      (im_height, im_width, 3)).astype(np.uint8)

TEST_IMAGE_PATHS = [ os.path.join(PATH_TO_TEST_IMAGES_DIR,'{}'.format(file)) for file in os.listdir(PATH_TO_TEST_IMAGES_DIR)]

print(TEST_IMAGE_PATHS)


#TEST_IMAGE_PATHS = [ os.path.join(PATH_TO_TEST_IMAGES_DIR, 'image#{}.jpg'.format(i)) for i in range(1, 3) ]

# Size, in inches, of the output images.
IMAGE_SIZE = (12, 8)

def write_jpeg(data, filepath):
    g = tf.Graph()
    with g.as_default():
        data_t = tf.placeholder(tf.uint8)
        op = tf.image.encode_jpeg(data_t, format='rgb', quality=100)
        init = tf.initialize_all_variables()

    with tf.Session(graph=g) as sess:
      sess.run(init)
      data_np = sess.run(op, feed_dict={ data_t: data })

    with open(filepath, 'wb') as fd:
        fd.write(data_np)

with detection_graph.as_default():
  with tf.Session(graph=detection_graph) as sess:
    # Definite input and output Tensors for detection_graph
    image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
    # Each box represents a part of the image where a particular object was detected.
    detection_boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
    # Each score represent how level of confidence for each of the objects.
    # Score is shown on the result image, together with the class label.
    detection_scores = detection_graph.get_tensor_by_name('detection_scores:0')
    detection_classes = detection_graph.get_tensor_by_name('detection_classes:0')
    num_detections = detection_graph.get_tensor_by_name('num_detections:0')
    for image_path in TEST_IMAGE_PATHS:
      image = Image.open(image_path)
      # the array based representation of the image will be used later in order to prepare the
      # result image with boxes and labels on it.
      image_np = load_image_into_numpy_array(image)
      # Expand dimensions since the model expects images to have shape: [1, None, None, 3]
      image_np_expanded = np.expand_dims(image_np, axis=0)
      # Actual detection.
      (boxes, scores, classes, num) = sess.run(
          [detection_boxes, detection_scores, detection_classes, num_detections],
          feed_dict={image_tensor: image_np_expanded})
      # Visualization of the results of a detection.
      vis_util.visualize_boxes_and_labels_on_image_array(
          image_np,
          np.squeeze(boxes),
          np.squeeze(classes).astype(np.int32),
          np.squeeze(scores),
          category_index,
          use_normalized_coordinates=True,
          line_thickness=8)
      write_jpeg(image_np, os.path.join(os.path.dirname(image_path),'inferred', os.path.basename(image_path)))
      plt.figure(figsize=IMAGE_SIZE)
      plt.imshow(image_np)
      plt.show()
      print(image_path)
      print(boxes)
      print(classes)
      print(scores)
      #write_jpeg(image_np, os.path.join(os.path.dirname(image_path),os.path.splitext(os.path.basename(image_path))[1]))

---

检测失败，因为培训需要很多步骤。我从2000步开始得到结果。jpeg损坏错误可能与牛津数据集有关，但我仍然能够使用该错误进行测试

---

哪个文件会导致此错误？你能手动打开吗？当我浏览它们时，所有的图像文件看起来都很好。我认为这个错误是随机的。评估在另一个检查点之后再次开始&错误重复。我的问题有两个。一个是评估，另一个是检测。即使评估失败，当我使用训练集中的图像时，检测也应该成功。但我甚至没有一个边界框。。。