加載一次YOLOv3預訓練模型即可用於對圖片預測，而多次加載預訓練模型會導致GPU內存不夠用。
2019-08-18 14:18:37.202250: I tensorflow/core/common_runtime/gpu/gpu_device.cc:993] Creating TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 1587 MB memory) -> physical GPU (device: 0, name: GeForce 930M, pci bus id: 0000:04:00.0, compute capability: 5.0)
2019-08-18 14:18:45.700148: I tensorflow/core/common_runtime/gpu/gpu_device.cc:993] Creating TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 107 MB memory) -> physical GPU (device: 0, name: GeForce 930M, pci bus id: 0000:04:00.0, compute capability: 5.0)
需要修改如下的代碼（多次加載預訓練模型）：
img_list = os.listdir(picture_dir)
for i in range(0,len(img_list)):
img_path = os.path.join(picture_dir,img_list[i])
print(img_path)
img_ori = cv2.imread(img_path)
with tf.Session() as sess:
yolo_model = yolov3(args.num_class, args.anchors)
with tf.variable_scope('yolov3'):
pred_feature_maps = yolo_model.forward(input_data, False, False)
pred_boxes, pred_confs, pred_probs = yolo_model.predict(pred_feature_maps)
pred_scores = pred_confs * pred_probs
boxes, scores, labels = gpu_nms(pred_boxes, pred_scores, args.num_class, max_boxes=200, score_thresh=0.3, nms_thresh=0.45)
saver = tf.train.Saver()
saver.restore(sess, args.restore_path)
boxes_, scores_, labels_ = sess.run([boxes, scores, labels], feed_dict={input_data: img})
wu@wu-X555LF:~/YOLOv3_TensorFlow-master$ python predict1.py
2019-08-18 14:18:35.219594: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2019-08-18 14:18:35.330784: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:898] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-18 14:18:35.331240: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1212] Found device 0 with properties:
name: GeForce 930M major: 5 minor: 0 memoryClockRate(GHz): 0.941
pciBusID: 0000:04:00.0
totalMemory: 1.96GiB freeMemory: 1.82GiB
2019-08-18 14:18:35.331263: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1312] Adding visible gpu devices: 0
2019-08-18 14:18:37.202250: I tensorflow/core/common_runtime/gpu/gpu_device.cc:993] Creating TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 1587 MB memory) -> physical GPU (device: 0, name: GeForce 930M, pci bus id: 0000:04:00.0, compute capability: 5.0)
box coords:
[]
******************************
scores:
[]
******************************
labels:
[]
2019-08-18 14:18:45.699982: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1312] Adding visible gpu devices: 0
2019-08-18 14:18:45.700148: I tensorflow/core/common_runtime/gpu/gpu_device.cc:993] Creating TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 107 MB memory) -> physical GPU (device: 0, name: GeForce 930M, pci bus id: 0000:04:00.0, compute capability: 5.0)
Traceback (most recent call last):
File "predict1.py", line 66, in <module>
pred_feature_maps = yolo_model.forward(input_data, False, False)
File "/home/wu/YOLOv3_TensorFlow-master/model.py", line 51, in forward
route_1, route_2, route_3 = darknet53_body(inputs)
File "/home/wu/YOLOv3_TensorFlow-master/utils/layer_utils.py", line 35, in darknet53_body
net = conv2d(inputs, 32, 3, strides=1)
File "/home/wu/YOLOv3_TensorFlow-master/utils/layer_utils.py", line 21, in conv2d
padding=('SAME' if strides == 1 else 'VALID'))
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/framework/python/ops/arg_scope.py", line 183, in func_with_args
return func(*args, **current_args)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/layers/python/layers/layers.py", line 1050, in convolution
outputs = layer.apply(inputs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/layers/base.py", line 809, in apply
return self.__call__(inputs, *args, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/layers/base.py", line 680, in __call__
self.build(input_shapes)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/layers/convolutional.py", line 143, in build
dtype=self.dtype)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/layers/base.py", line 533, in add_variable
partitioner=partitioner)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 1297, in get_variable
constraint=constraint)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 1093, in get_variable
constraint=constraint)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 431, in get_variable
return custom_getter(**custom_getter_kwargs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/layers/python/layers/layers.py", line 1613, in layer_variable_getter
return _model_variable_getter(getter, *args, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/layers/python/layers/layers.py", line 1604, in _model_variable_getter
use_resource=use_resource)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/framework/python/ops/arg_scope.py", line 183, in func_with_args
return func(*args, **current_args)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/framework/python/ops/variables.py", line 291, in model_variable
use_resource=use_resource)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/framework/python/ops/arg_scope.py", line 183, in func_with_args
return func(*args, **current_args)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/framework/python/ops/variables.py", line 246, in variable
use_resource=use_resource)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 408, in _true_getter
use_resource=use_resource, constraint=constraint)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 747, in _get_single_variable
name, "".join(traceback.format_list(tb))))
ValueError: Variable yolov3/darknet53_body/Conv/weights already exists, disallowed. Did you mean to set reuse=True or reuse=tf.AUTO_REUSE in VarScope? Originally defined at:
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/framework/python/ops/variables.py", line 246, in variable
use_resource=use_resource)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/framework/python/ops/arg_scope.py", line 183, in func_with_args
return func(*args, **current_args)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/framework/python/ops/variables.py", line 291, in model_variable
use_resource=use_resource)
wu@wu-X555LF:~/YOLOv3_TensorFlow-master$
修改爲（加載一次YOLOv3預訓練模型用於對所有圖片預測）：
with tf.Session() as sess:
yolo_model = yolov3(args.num_class, args.anchors)
with tf.variable_scope('yolov3'):
pred_feature_maps = yolo_model.forward(input_data, False)
pred_boxes, pred_confs, pred_probs = yolo_model.predict(pred_feature_maps)
pred_scores = pred_confs * pred_probs
boxes, scores, labels = gpu_nms(pred_boxes, pred_scores, args.num_class, max_boxes=200, score_thresh=0.3, nms_thresh=0.45)
saver = tf.train.Saver()
saver.restore(sess, args.restore_path)
img_list = os.listdir(picture_dir)
for i in range(0,len(img_list)):
img_path = os.path.join(picture_dir,img_list[i])
img_ori = cv2.imread(img_path) #args.input_image
if args.letterbox_resize:
img, resize_ratio, dw, dh = letterbox_resize(img_ori, args.new_size[0], args.new_size[1])
else:
height_ori, width_ori = img_ori.shape[:2]
boxes_, scores_, labels_ = sess.run([boxes, scores, labels], feed_dict={input_data: img})
# coding: utf-8
from __future__ import division, print_function
import tensorflow as tf
import numpy as np
import argparse
import cv2
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
from utils.misc_utils import parse_anchors, read_class_names
from utils.nms_utils import gpu_nms
from utils.plot_utils import get_color_table, plot_one_box
from utils.data_aug import letterbox_resize
from model import yolov3
picture_dir = './data/demo_data'
output_dir = './predict.txt'
output_txt = open(output_dir,"w");
parser = argparse.ArgumentParser(description="YOLO-V3 test single image test procedure.")
#parser.add_argument("input_image", type=str,
# help="The path of the input image.")
parser.add_argument("--anchor_path", type=str, default="./data/yolo_anchors.txt",
help="The path of the anchor txt file.")
parser.add_argument("--new_size", nargs='*', type=int, default=[416, 416],
help="Resize the input image with `new_size`, size format: [width, height]")
parser.add_argument("--letterbox_resize", type=lambda x: (str(x).lower() == 'true'), default=True,
help="Whether to use the letterbox resize.")
parser.add_argument("--class_name_path", type=str, default="./data/my_data/data.names",
help="The path of the class names.")
parser.add_argument("--restore_path", type=str, default="./data/darknet_weights/best/best_model_Epoch_98_step_50291_mAP_0.7358_loss_3.6063_lr_1e-05",
help="The path of the weights to restore.")
args = parser.parse_args()
args.anchors = parse_anchors(args.anchor_path)
args.classes = read_class_names(args.class_name_path)
args.num_class = len(args.classes)
color_table = get_color_table(args.num_class)
with tf.Session() as sess:
input_data = tf.placeholder(tf.float32, [1, args.new_size[1], args.new_size[0], 3], name='input_data')
yolo_model = yolov3(args.num_class, args.anchors)
with tf.variable_scope('yolov3'):
pred_feature_maps = yolo_model.forward(input_data, False)
pred_boxes, pred_confs, pred_probs = yolo_model.predict(pred_feature_maps)
pred_scores = pred_confs * pred_probs
boxes, scores, labels = gpu_nms(pred_boxes, pred_scores, args.num_class, max_boxes=200, score_thresh=0.3, nms_thresh=0.45)
saver = tf.train.Saver()
saver.restore(sess, args.restore_path)
img_list = os.listdir(picture_dir)
for i in range(0,len(img_list)):
img_path = os.path.join(picture_dir,img_list[i])
img_ori = cv2.imread(img_path) #args.input_image
if args.letterbox_resize:
img, resize_ratio, dw, dh = letterbox_resize(img_ori, args.new_size[0], args.new_size[1])
else:
height_ori, width_ori = img_ori.shape[:2]
img = cv2.resize(img_ori, tuple(args.new_size))
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img = np.asarray(img, np.float32)
img = img[np.newaxis, :] / 255.
boxes_, scores_, labels_ = sess.run([boxes, scores, labels], feed_dict={input_data: img})
# rescale the coordinates to the original image
if args.letterbox_resize:
boxes_[:, [0, 2]] = (boxes_[:, [0, 2]] - dw) / resize_ratio
boxes_[:, [1, 3]] = (boxes_[:, [1, 3]] - dh) / resize_ratio
else:
boxes_[:, [0, 2]] *= (width_ori/float(args.new_size[0]))
boxes_[:, [1, 3]] *= (height_ori/float(args.new_size[1]))
print("box coords:")
print(boxes_)
print('*' * 30)
print("scores:")
print(scores_)
print('*' * 30)
print("labels:")
print(labels_)
output_txt.write(img_path) #ig_path
for j in range(len(boxes_)):
label_class = ' '+ str(labels_[j]) + ' '
x0, y0, x1, y1 = boxes_[j]
locate = str(x0)+' '+str(y0)+' '+str(x1)+' '+str(y1)
output_txt.write(label_class)
output_txt.write(locate)
output_txt.write('\n')
#print(type(labels_[j])) <type 'numpy.int32'>
#print(type(boxes_[j])) <type 'numpy.ndarray'>
#output_txt.write(labels_[i]) #classes_label
#output_txt.write(boxes_[j]) #[x0, y0, x1, y1]
#plot_one_box(img_ori, [x0, y0, x1, y1], label=args.classes[labels_[i]] + ', {:.2f}%'.format(scores_[i] * 100), color=color_table[labels_[i]])
#cv2.imshow('Detection result', img_ori)
#cv2.imwrite('detection_result.jpg', img_ori)
#cv2.waitKey(0)
預測結果：
wu@wu-X555LF:~/YOLOv3_TensorFlow-master$ python predict_add.py
2019-08-18 14:33:33.506594: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2019-08-18 14:33:33.583756: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:898] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-18 14:33:33.584216: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1212] Found device 0 with properties:
name: GeForce 930M major: 5 minor: 0 memoryClockRate(GHz): 0.941
pciBusID: 0000:04:00.0
totalMemory: 1.96GiB freeMemory: 1.84GiB
2019-08-18 14:33:33.584248: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1312] Adding visible gpu devices: 0
2019-08-18 14:33:34.217678: I tensorflow/core/common_runtime/gpu/gpu_device.cc:993] Creating TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 1606 MB memory) -> physical GPU (device: 0, name: GeForce 930M, pci bus id: 0000:04:00.0, compute capability: 5.0)
box coords:
[[-195.83458 163.17838 884.7618 1150.6211 ]
[ -31.841726 391.56403 702.5625 912.8325 ]]
******************************
scores:
[0.9449485 0.7733519]
******************************
labels:
[2 2]
box coords:
[]
******************************
scores:
[]
******************************
labels:
[]
box coords:
[[314.3078 468.14548 390.6185 567.1601 ]
[-41.491894 514.53687 213.59131 609.0764 ]
[ 15.074331 462.53992 156.97115 728.9914 ]
[-41.491894 514.53687 213.59131 609.0764 ]
[ 15.074331 462.53992 156.97115 728.9914 ]
[229.68753 852.45557 244.8793 954.3424 ]
[135.6899 759.6658 154.38506 844.6287 ]
[133.06404 731.6753 143.916 811.71185 ]
[217.71773 875.35455 236.87387 987.4167 ]]
******************************
scores:
[0.3414791 0.45428857 0.31732422 0.4882345 0.3644903 0.5662823
0.46963194 0.32721734 0.3194868 ]
******************************
labels:
[1 4 4 7 7 9 9 9 9]
box coords:
[[110.1114 56.36794 241.52951 316.10345 ]
[241.23412 49.088387 326.56253 264.94296 ]
[131.32816 86.27491 210.68886 273.31342 ]]
******************************
scores:
[0.68328035 0.44021612 0.4340352 ]
******************************
labels:
[0 0 0]
box coords:
[]
******************************
scores:
[]
******************************
labels:
[]
box coords:
[[265.07657 77.91465 331.14487 305.70837 ]
[412.98926 76.29398 464.5738 350.3297 ]
[ 37.823307 31.951302 265.8799 321.9834 ]
[127.45379 25.41013 214.79816 325.07767 ]
[387.29068 72.292145 440.42075 344.67874 ]]
******************************
scores:
[0.82978487 0.64961237 0.5234746 0.44627655 0.3882754 ]
******************************
labels:
[0 0 0 0 0]
wu@wu-X555LF:~/YOLOv3_TensorFlow-master$