VSAG/vision_sjtu_19
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

修改了摄像头读取方式

Browse Source
This commit is contained in:
xinyang
2019-04-27 16:16:53 +08:00
parent 9cfd26cc23
commit 4e47b38d7d
15 changed files with 272 additions and 255 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
tools/TrainCNN/backward.py
View File

@@ -2,7 +2,8 @@ import tensorflow as tf
from progressive.bar import Bar
import generate
import forward
import cv2
import numpy as np
def save_kernal(fp, val):
print(val.shape[2], file=fp)
@@ -49,7 +50,7 @@ def save_para(folder, paras):
save_bias(fp, paras[7])
STEPS = 30000
STEPS = 20000
BATCH = 10
LEARNING_RATE_BASE = 0.01
LEARNING_RATE_DECAY = 0.99
@@ -59,9 +60,9 @@ MOVING_AVERAGE_DECAY = 0.99
def train(dataset, show_bar=False):
test_images, test_labels = dataset.all_test_sets()
x = tf.placeholder(tf.float32, [None, forward.SRC_ROWS, forward.SRC_COLS, forward.SRC_CHANNELS])
x = tf.placeholder(tf.float32, [None, generate.SRC_ROWS, generate.SRC_COLS, generate.SRC_CHANNELS])
y_= tf.placeholder(tf.float32, [None, forward.OUTPUT_NODES])
nodes, vars = forward.forward(0.001)
nodes, vars = forward.forward(x, 0.001)
y = nodes[-1]
ce = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_, 1))
@@ -72,7 +73,7 @@ def train(dataset, show_bar=False):
learning_rate = tf.train.exponential_decay(
LEARNING_RATE_BASE,
global_step,
len(dataset.train_sets) / BATCH,
len(dataset.train_samples) / BATCH,
LEARNING_RATE_DECAY,
staircase=False)
train_step = tf.train.AdamOptimizer(learning_rate).minimize(loss, global_step=global_step)
@@ -112,12 +113,31 @@ def train(dataset, show_bar=False):
bar.cursor.restore()
bar.draw(value=i+1)
# video = cv2.VideoCapture("/home/xinyang/Desktop/Video.mp4")
# _ = True
# while _:
# _, frame = video.read()
# cv2.imshow("Video", frame)
# if cv2.waitKey(10) == 113:
# bbox = cv2.selectROI("frame", frame, False)
# print(bbox)
# roi = frame[bbox[1]:bbox[1]+bbox[3], bbox[0]:bbox[0]+bbox[2]]
# roi = cv2.resize(roi, (48, 36))
# cv2.imshow("roi", roi)
# cv2.waitKey(0)
# roi = roi.astype(np.float32)
# roi /= 255.0
# roi = roi.reshape([1, 36, 48, 3])
# res = sess.run(y, feed_dict={x: roi})
# res = res.reshape([forward.OUTPUT_NODES])
# print(np.argmax(res))
vars_val = sess.run(vars)
save_para("paras", vars_val)
# nodes_val = sess.run(nodes, feed_dict={x:test})
# return vars_val, nodes_val
save_para("/home/xinyang/Desktop/AutoAim/tools/para", vars_val)
nodes_val = sess.run(nodes, feed_dict={x:test_images})
return vars_val, nodes_val
if __name__ == "__main__":
dataset = generate.DataSet("images")
dataset = generate.DataSet("/home/xinyang/Desktop/DataSets")
train(dataset, show_bar=True)

21
tools/TrainCNN/forward.py
View File

@@ -3,7 +3,7 @@ import tensorflow as tf
def get_weight(shape, regularizer=None):
w = tf.Variable(tf.truncated_normal(shape, stddev=0.1))
if regularizer is None:
if regularizer is not None:
tf.add_to_collection('losses', tf.contrib.layers.l2_regularizer(regularizer)(w))
return w
@@ -25,32 +25,23 @@ def max_pool_2x2(x):
return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="VALID")
# 原图像行数
SRC_ROWS = 36
# 原图像列数
SRC_COLS = 48
# 原图像通道数
SRC_CHANNELS = 1
# 第一层卷积核大小
CONV1_KERNAL_SIZE = 5
# 第一层卷积输出通道数
CONV1_OUTPUT_CHANNELS = 8
CONV1_OUTPUT_CHANNELS = 4
# 第二层卷积核大小
CONV2_KERNAL_SIZE = 3
# 第二层卷积输出通道数
CONV2_OUTPUT_CHANNELS = 16
CONV2_OUTPUT_CHANNELS = 8
# 第一层全连接宽度
FC1_OUTPUT_NODES = 32
FC1_OUTPUT_NODES = 16
# 第二层全连接宽度（输出标签类型数）
FC2_OUTPUT_NODES = 8
FC2_OUTPUT_NODES = 4
# 输出标签类型数
OUTPUT_NODES = FC2_OUTPUT_NODES
@@ -61,7 +52,7 @@ def forward(x, regularizer=None):
nodes = []
conv1_w = get_weight(
[CONV1_KERNAL_SIZE, CONV1_KERNAL_SIZE, 1, CONV1_OUTPUT_CHANNELS]
[CONV1_KERNAL_SIZE, CONV1_KERNAL_SIZE, int(x.shape[3]), CONV1_OUTPUT_CHANNELS]
)
conv1_b = get_bias([CONV1_OUTPUT_CHANNELS])
conv1 = tf.nn.relu(tf.nn.bias_add(conv2d(x, conv1_w), conv1_b))

60
tools/TrainCNN/generate.py
View File

@@ -3,47 +3,63 @@ import os
import cv2
import random
from forward import OUTPUT_NODES
# 原图像行数
SRC_ROWS = 36
# 原图像列数
SRC_COLS = 48
# 原图像通道数
SRC_CHANNELS = 3
class DataSet:
def __init__(self, folder):
self.train_sets = []
self.test_sets = []
self.train_samples = []
self.train_labels = []
self.test_samples = []
self.test_labels = []
self.generate_data_sets(folder)
def file2nparray(self, name):
image = cv2.imread(name)
image = cv2.resize(image, (SRC_COLS, SRC_ROWS))
image = image.astype(np.float32)
return image / 255.0
def id2label(self, id):
a = np.zeros([OUTPUT_NODES])
a[id] = 1
return a[:]
def generate_data_sets(self, folder):
def file2nparray(name):
image = cv2.imread(name)
return image[:, :, 0]
def id2label(id):
a = np.zeros([OUTPUT_NODES, 1])
a[id] = 1
return a[:]
sets = []
for i in range(OUTPUT_NODES):
dir = "%s/%d" % (folder, i)
files = os.listdir(dir)
for file in files:
sets.append([file2nparray("%s/%s" % (dir, file)), id2label(i)])
sets = np.array(sets)
np.random.shuffle(sets)
length = len(sets)
self.train_sets = sets[:length*3//4]
self.test_sets = sets[length*3//4:]
if random.random() > 0.2:
self.train_samples.append(self.file2nparray("%s/%s" % (dir, file)))
self.train_labels.append(self.id2label(i))
else:
self.test_samples.append(self.file2nparray("%s/%s" % (dir, file)))
self.test_labels.append(self.id2label(i))
self.train_samples = np.array(self.train_samples)
self.train_labels = np.array(self.train_labels)
self.test_samples = np.array(self.test_samples)
self.test_labels = np.array(self.test_labels)
return sets
def sample_train_sets(self, length):
samples = []
labels = []
for i in range(length):
id = random.randint(0, length-1)
samples.append(self.train_sets[id][0])
labels.append(self.train_sets[id][1])
id = random.randint(0, len(self.train_samples)-1)
samples.append(self.train_samples[id])
labels.append(self.train_labels[id])
return np.array(samples), np.array(labels)
def all_train_sets(self):
return self.train_sets[:, 0, :, :], self.train_sets[:, 1, :, :]
return self.train_samples[:], self.train_labels[:]
def all_test_sets(self):
return self.test_sets[:, 0, :, :], self.test_sets[:, 1, :, :]
return self.test_samples[:], self.test_labels[:]