def main(checkpoint, input_files):
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
model_config, _, track_config = load_cfgs(checkpoint)
track_config['log_level'] = 1
track_config["is_video"] = False
g = tf.Graph()
with g.as_default():
model = inference_wrapper.InferenceWrapper()
restore_fn = model.build_graph_from_config(model_config, track_config,
checkpoint)
g.finalize()
if not osp.isdir(track_config['log_dir']):
logging.info('Creating inference directory: %s',
track_config['log_dir'])
mkdir_p(track_config['log_dir'])
video_dirs = []
for file_pattern in input_files.split(","):
video_dirs.extend(glob(file_pattern))
logging.info("Running tracking on %d videos matching %s", len(video_dirs),
input_files)
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
with tf.Session(graph=g, config=sess_config) as sess:
restore_fn(sess)
tracker = Tracker(model,
model_config=model_config,
track_config=track_config)
for video_dir in video_dirs:
if not osp.isdir(video_dir):
logging.warning(
'{} is not a directory, skipping...'.format(video_dir))
continue
video_name = osp.basename(video_dir)
video_log_dir = osp.join(track_config['log_dir'], video_name)
mkdir_p(video_log_dir)
filenames = sort_nicely(glob(video_dir + '/img/*.jpg'))
first_line = open(video_dir + '/groundtruth_rect.txt').readline()
bb = [int(v) for v in first_line.strip().split(',')]
# Rectangle: [x,y,width,height]
init_bb = Rectangle(bb[0] - 1, bb[1] - 1, bb[2],
bb[3]) # 0-index in python
trajectory = tracker.track(sess, init_bb, filenames, video_log_dir)
with open(osp.join(video_log_dir, 'track_rect.txt'), 'w') as f:
for region in trajectory:
rect_str = '{},{},{},{}\n'.format(region.x + 1,
region.y + 1,
region.width,
region.height)
f.write(rect_str)
def run_SiamRPN_OPF(seq, rp, bSaveImage):
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
config_name = "SiamRPN_ftall"
CHECKPOINT = '/home/lab-xiong.jiangfeng/Projects/SiameseRPN/Logs/%s/track_model_checkpoints/%s' % (
config_name, config_name)
logging.info('Evaluating {}...'.format(CHECKPOINT))
# Read configurations from json
model_config, _, track_config = load_cfgs(CHECKPOINT)
track_config['log_level'] = 0 # Skip verbose logging for speed
np.random.seed(1234)
tf.set_random_seed(1234)
g = tf.Graph()
with g.as_default():
model = get_model(model_config['Model'])(model_config=model_config,
mode='inference')
model.build(reuse=tf.AUTO_REUSE)
model.online_net = OnlineNet(online_config,
is_training=True,
reuse=False)
model.online_valnet = OnlineNet(online_config,
is_training=False,
reuse=True)
global_variables_init_op = tf.global_variables_initializer()
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
sess_config.gpu_options.per_process_gpu_memory_fraction = 0.2
with tf.Session(graph=g, config=sess_config) as sess:
sess.run(global_variables_init_op)
model.restore_weights_from_checkpoint(sess, 605000)
tracker = OnlineTracker(sess,
model,
track_config,
online_config,
show_video=0)
tic = time.clock()
frames = seq.s_frames
init_rect = seq.init_rect
x, y, width, height = init_rect # OTB format
init_bb = Rectangle(x - 1, y - 1, width, height)
trajectory_py = tracker.track(init_bb, frames, bSaveImage, rp)
#print(trajectory_py)
trajectory = [
Rectangle(val.x + 1, val.y + 1, val.width, val.height)
for val in trajectory_py
] # x, y add one to match OTB format
duration = time.clock() - tic
result = dict()
result['res'] = trajectory
result['type'] = 'rect'
result['fps'] = round(seq.len / duration, 3)
return result
def main(checkpoint, input_files):
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
model_config, _, track_config = load_cfgs(checkpoint)
track_config['log_level'] = 1
g = tf.Graph()
with g.as_default():
model = inference_wrapper.InferenceWrapper()
restore_fn = model.build_graph_from_config(model_config, track_config, checkpoint)
g.finalize()
if not osp.isdir(track_config['log_dir']):
logging.info('Creating inference directory: %s', track_config['log_dir'])
mkdir_p(track_config['log_dir'])
video_dirs = []
for file_pattern in input_files.split(","):
video_dirs.extend(glob(file_pattern))
logging.info("Running tracking on %d videos matching %s", len(video_dirs), input_files)
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
with tf.Session(graph=g, config=sess_config) as sess:
restore_fn(sess)
tracker = Tracker(model, model_config=model_config, track_config=track_config)
for video_dir in video_dirs:
if not osp.isdir(video_dir):
logging.warning('{} is not a directory, skipping...'.format(video_dir))
continue
video_name = osp.basename(video_dir)
video_log_dir = osp.join(track_config['log_dir'], video_name)
mkdir_p(video_log_dir)
filenames = sort_nicely(glob(video_dir + '/img/*.jpg'))
first_line = open(video_dir + '/groundtruth_rect.txt').readline()
bb = [int(v) for v in first_line.strip().split(',')]
init_bb = Rectangle(bb[0] - 1, bb[1] - 1, bb[2], bb[3]) # 0-index in python
trajectory = tracker.track(sess, init_bb, filenames, video_log_dir)
with open(osp.join(video_log_dir, 'track_rect.txt'), 'w') as f:
for region in trajectory:
rect_str = '{},{},{},{}\n'.format(region.x + 1, region.y + 1,
region.width, region.height)
f.write(rect_str)
def __init__(
self,
debug=0,
checkpoint='Logs/SiamFC/track_model_checkpoints/SiamFC-3s-color-pretrained'
):
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
# run only on cpu
# os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
model_config, _, track_config = load_cfgs(checkpoint)
track_config['log_level'] = debug
g = tf.Graph()
with g.as_default():
model = inference_wrapper.InferenceWrapper()
restore_fn = model.build_graph_from_config(model_config,
track_config,
checkpoint)
g.finalize()
if not osp.isdir(track_config['log_dir']):
logging.info('Creating inference directory: %s',
track_config['log_dir'])
mkdir_p(track_config['log_dir'])
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
sess = tf.Session(graph=g, config=sess_config)
# sess.run(tf.global_variables_initializer())
restore_fn(sess)
tracker = Tracker(model,
model_config=model_config,
track_config=track_config)
video_name = "demo"
video_log_dir = osp.join(track_config['log_dir'], video_name)
rmdir(video_log_dir)
mkdir_p(video_log_dir)
self.tracker = tracker
self.sess = sess
self.video_log_dir = video_log_dir
self.graph = g
import os import os.path as osp import sys import numpy as np import tensorflow as tf CURRENT_DIR = osp.dirname(__file__) sys.path.append(osp.join(CURRENT_DIR, '..')) import configuration import siamese_model from utils.misc_utils import auto_select_gpu, save_cfgs # Set GPU os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu() tf.logging.set_verbosity(tf.logging.DEBUG) from sacred import Experiment ex = Experiment(configuration.RUN_NAME) @ex.config def configurations(): # Add configurations for current script, for more details please see the documentation of `sacred`. model_config = configuration.MODEL_CONFIG train_config = configuration.TRAIN_CONFIG track_config = configuration.TRACK_CONFIG
def main(model_config, train_config, track_config):
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
# Create training directory which will be used to save: configurations, model files, TensorBoard logs
train_dir = train_config['train_dir']
if not osp.isdir(train_dir):
logging.info('Creating training directory: %s', train_dir)
mkdir_p(train_dir)
g = tf.Graph()
with g.as_default():
# Set fixed seed for reproducible experiments
random.seed(train_config['seed'])
np.random.seed(train_config['seed'])
tf.set_random_seed(train_config['seed'])
# Build the training and validation model
model = siamese_model.SiameseModel(model_config,
train_config,
mode='train')
model.build()
model_va = siamese_model.SiameseModel(model_config,
train_config,
mode='validation')
model_va.build(reuse=True)
# Save configurations for future reference
save_cfgs(train_dir, model_config, train_config, track_config)
learning_rate = _configure_learning_rate(train_config,
model.global_step)
optimizer = _configure_optimizer(train_config, learning_rate)
tf.summary.scalar('learning_rate', learning_rate)
# Set up the training ops
opt_op = tf.contrib.layers.optimize_loss(
loss=model.total_loss,
global_step=model.global_step,
learning_rate=learning_rate,
optimizer=optimizer,
clip_gradients=train_config['clip_gradients'],
learning_rate_decay_fn=None,
summaries=['learning_rate'])
with tf.control_dependencies([opt_op]):
train_op = tf.no_op(name='train')
saver = tf.train.Saver(
tf.global_variables(),
max_to_keep=train_config['max_checkpoints_to_keep'])
summary_writer = tf.summary.FileWriter(train_dir, g)
summary_op = tf.summary.merge_all()
global_variables_init_op = tf.global_variables_initializer()
local_variables_init_op = tf.local_variables_initializer()
g.finalize() # Finalize graph to avoid adding ops by mistake
# Dynamically allocate GPU memory
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
sess = tf.Session(config=sess_config)
model_path = tf.train.latest_checkpoint(train_config['train_dir'])
if not model_path:
sess.run(global_variables_init_op)
sess.run(local_variables_init_op)
start_step = 0
if model_config['embed_config']['embedding_checkpoint_file']:
model.init_fn(sess)
else:
logging.info('Restore from last checkpoint: {}'.format(model_path))
sess.run(local_variables_init_op)
saver.restore(sess, model_path)
start_step = tf.train.global_step(sess, model.global_step.name) + 1
# Training loop
data_config = train_config['train_data_config']
total_steps = int(data_config['epoch'] *
data_config['num_examples_per_epoch'] /
data_config['batch_size'])
logging.info('Train for {} steps'.format(total_steps))
for step in range(start_step, total_steps):
start_time = time.time()
_, loss, batch_loss = sess.run(
[train_op, model.total_loss, model.batch_loss])
duration = time.time() - start_time
if step % 10 == 0:
examples_per_sec = data_config['batch_size'] / float(duration)
time_remain = data_config['batch_size'] * (
total_steps - step) / examples_per_sec
m, s = divmod(time_remain, 60)
h, m = divmod(m, 60)
format_str = (
'%s: step %d, total loss = %.2f, batch loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch; %dh:%02dm:%02ds remains)')
logging.info(format_str %
(datetime.now(), step, loss, batch_loss,
examples_per_sec, duration, h, m, s))
if step % 100 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
if step % train_config['save_model_every_n_step'] == 0 or (
step + 1) == total_steps:
checkpoint_path = osp.join(train_config['train_dir'],
'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
def main(model_config, train_config, track_config):
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
# Create training directory which will be used to save: configurations, model files, TensorBoard logs
train_dir = train_config['train_dir']
if not osp.isdir(train_dir):
logging.info('Creating training directory: %s', train_dir)
mkdir_p(train_dir)
if have_cfgs(train_dir):
model_config, train_config, track_config = load_cfgs(train_dir)
print("=================== load cfg ")
else:
save_cfgs(train_dir, model_config, train_config, track_config)
print("=================== save default cfg, please modify files in {}".format(train_dir))
return
g = tf.Graph()
with g.as_default():
# Set fixed seed for reproducible experiments
random.seed(train_config['seed'])
np.random.seed(train_config['seed'])
tf.set_random_seed(train_config['seed'])
# Build the training and validation model
model = siamese_model.SiameseModel(model_config, train_config, track_config, mode='train')
model.build()
model_va = siamese_model.SiameseModel(model_config, train_config, track_config, mode='validation')
model_va.build(reuse=True)
learning_rate = _configure_learning_rate(train_config, model.global_step)
optimizer = _configure_optimizer(train_config, learning_rate)
tf.summary.scalar('learning_rate', learning_rate)
# general way for run train: https://qiita.com/horiem/items/00ec6488b23895cc4fe2
# tensorflow 2.1: https://www.tensorflow.org/tutorials/customization/custom_training_walkthrough
# Set up the training ops
opt_op = tensorflow.contrib.layers.optimize_loss(
loss=model.total_loss,
global_step=model.global_step,
learning_rate=learning_rate,
optimizer=optimizer,
clip_gradients=train_config['clip_gradients'],
learning_rate_decay_fn=None,
summaries=['learning_rate'])
with tf.control_dependencies([opt_op]):
train_op = tf.no_op(name='train')
saver = tf.train.Saver(tf.global_variables(),
max_to_keep=train_config['max_checkpoints_to_keep'])
summary_writer = tf.summary.FileWriter(train_dir, g)
summary_op = tf.summary.merge_all()
global_variables_init_op = tf.global_variables_initializer()
local_variables_init_op = tf.local_variables_initializer()
g.finalize() # Finalize graph to avoid adding ops by mistake
# Dynamically allocate GPU memory
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
sess = tf.Session(config=sess_config)
model_path = tf.train.latest_checkpoint(train_config['train_dir'])
if not model_path:
sess.run(global_variables_init_op)
sess.run(local_variables_init_op)
start_step = 0
if model_config['embed_config']['embedding_checkpoint_file']:
model.init_fn(sess)
else:
logging.info('Restore from last checkpoint: {}'.format(model_path))
sess.run(local_variables_init_op)
saver.restore(sess, model_path)
start_step = tf.train.global_step(sess, model.global_step.name) + 1
# export
if train_config["export"]:
# still debugging
'''
frozen_graph_def = tf.graph_util.convert_variables_to_constants(sess, tf.get_default_graph().as_graph_def(), ["train/detection/add"])
frozen_graph = tf.Graph()
with frozen_graph.as_default():
tf.import_graph_def(frozen_graph_def)
save_model_dir = osp.join(train_config['train_dir'], 'models')
tf.train.write_graph(frozen_graph_def, save_model_dir, 'quantized_frozen_graph.pb', as_text=False)
tf.train.write_graph(frozen_graph_def, save_model_dir, 'quantized_frozen_graph.pbtxt', as_text=True)
output_op = sess.graph.get_tensor_by_name("validation/detection/add:0")
input1_op = sess.graph.get_tensor_by_name("validation/template_image:0")
input2_op = sess.graph.get_tensor_by_name("validation/input_image:0")
converter = tf.lite.TFLiteConverter.from_session(sess, [input1_op, input2_op], [output_op])
converter.inference_type = tf.lite.constants.QUANTIZED_UINT8
input_arrays = converter.get_input_arrays()
converter.quantized_input_stats = {input_arrays[0] : (0., 1.), input_arrays[1] : (0., 1.)} # mean, std_dev
converter.default_ranges_stats = (0, 255)
tflite_model = converter.convert()
open(osp.join(save_model_dir, 'quantized_frozen_graph.tflite'), "wb").write(tflite_model)
'''
return
# Training loop
data_config = train_config['train_data_config']
total_steps = int(data_config['epoch'] *
data_config['num_examples_per_epoch'] /
data_config['batch_size'])
logging.info('Train for {} steps'.format(total_steps))
save_step = int(data_config['num_examples_per_epoch'] / data_config['batch_size'])
print("=========== save_step: {}".format(save_step))
for step in range(start_step, total_steps):
start_time = time.time()
# no "feed_dict"
# has "feed_dict" exmaple (mnist): https://qiita.com/SwitchBlade/items/6677c283b2402d060cd0
_, loss, batch_loss, instances, response = sess.run([train_op, model.total_loss, model.batch_loss, model.instances, model.response])
duration = time.time() - start_time
if step % 10 == 0:
examples_per_sec = data_config['batch_size'] / float(duration)
time_remain = data_config['batch_size'] * (total_steps - step) / examples_per_sec
m, s = divmod(time_remain, 60)
h, m = divmod(m, 60)
format_str = ('%s: step %d, total loss = %.2f, batch loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch; %dh:%02dm:%02ds remains)')
logging.info(format_str % (datetime.now(), step, loss, batch_loss,
examples_per_sec, duration, h, m, s))
if step % 100 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
if step % save_step == 0 or (step + 1) == total_steps:
checkpoint_path = osp.join(train_config['train_dir'], 'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
def main(model_config, train_config, track_config):
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
# Create training directory which will be used to save: configurations, model files, TensorBoard logs
train_dir = train_config['train_dir']
if not osp.isdir(train_dir):
logging.info('Creating training directory: %s', train_dir)
mkdir_p(train_dir)
g = tf.Graph()
with g.as_default():
# Set fixed seed for reproducible experiments
random.seed(train_config['seed'])
np.random.seed(train_config['seed'])
tf.set_random_seed(train_config['seed'])
# Build the training and validation model
model = siamese_model.SiameseModel(model_config, train_config, mode='train')
model.build()
model_va = siamese_model.SiameseModel(model_config, train_config, mode='validation')
model_va.build(reuse=True)
# Save configurations for future reference
save_cfgs(train_dir, model_config, train_config, track_config)
learning_rate = _configure_learning_rate(train_config, model.global_step)
optimizer = _configure_optimizer(train_config, learning_rate)
tf.summary.scalar('learning_rate', learning_rate)
# Set up the training ops
opt_op = tf.contrib.layers.optimize_loss(
loss=model.total_loss,
global_step=model.global_step,
learning_rate=learning_rate,
optimizer=optimizer,
clip_gradients=train_config['clip_gradients'],
learning_rate_decay_fn=None,
summaries=['learning_rate'])
with tf.control_dependencies([opt_op]):
train_op = tf.no_op(name='train')
saver = tf.train.Saver(tf.global_variables(),
max_to_keep=train_config['max_checkpoints_to_keep'])
summary_writer = tf.summary.FileWriter(train_dir, g)
summary_op = tf.summary.merge_all()
global_variables_init_op = tf.global_variables_initializer()
local_variables_init_op = tf.local_variables_initializer()
g.finalize() # Finalize graph to avoid adding ops by mistake
# Dynamically allocate GPU memory
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
sess = tf.Session(config=sess_config)
model_path = tf.train.latest_checkpoint(train_config['train_dir'])
if not model_path:
sess.run(global_variables_init_op)
sess.run(local_variables_init_op)
start_step = 0
if model_config['embed_config']['embedding_checkpoint_file']:
model.init_fn(sess)
else:
logging.info('Restore from last checkpoint: {}'.format(model_path))
sess.run(local_variables_init_op)
saver.restore(sess, model_path)
start_step = tf.train.global_step(sess, model.global_step.name) + 1
# Training loop
data_config = train_config['train_data_config']
total_steps = int(data_config['epoch'] *
data_config['num_examples_per_epoch'] /
data_config['batch_size'])
logging.info('Train for {} steps'.format(total_steps))
for step in range(start_step, total_steps):
start_time = time.time()
_, loss, batch_loss = sess.run([train_op, model.total_loss, model.batch_loss])
duration = time.time() - start_time
if step % 10 == 0:
examples_per_sec = data_config['batch_size'] / float(duration)
time_remain = data_config['batch_size'] * (total_steps - step) / examples_per_sec
m, s = divmod(time_remain, 60)
h, m = divmod(m, 60)
format_str = ('%s: step %d, total loss = %.2f, batch loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch; %dh:%02dm:%02ds remains)')
logging.info(format_str % (datetime.now(), step, loss, batch_loss,
examples_per_sec, duration, h, m, s))
if step % 100 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
if step % train_config['save_model_every_n_step'] == 0 or (step + 1) == total_steps:
checkpoint_path = osp.join(train_config['train_dir'], 'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
def main(model_config, train_config, track_config):
# GPU Config
gpu_list = train_config['train_data_config'].get('gpu_ids', '0')
num_gpus = len(gpu_list.split(','))
if num_gpus > 1:
os.environ['CUDA_VISIBLE_DEVICES'] = gpu_list
else:
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
# Create training directory which will be used to save: configurations, model files, TensorBoard logs
train_dir = train_config['train_dir']
if not osp.isdir(train_dir):
logging.info('Creating training directory: %s', train_dir)
mkdir_p(train_dir)
g = tf.Graph()
with g.as_default():
# Set fixed seed for reproducible experiments
random.seed(train_config['seed'])
np.random.seed(train_config['seed'])
tf.set_random_seed(train_config['seed'])
#Build global step
with tf.name_scope('train/'):
global_step = tf.Variable(initial_value=0,
name='global_step',
trainable=False,
collections=[
tf.GraphKeys.GLOBAL_STEP,
tf.GraphKeys.GLOBAL_VARIABLES
])
Model = get_model(model_config['Model'])
# build training dataloader and validation dataloader
#---train
train_dataloader = DataLoader(train_config['train_data_config'],
is_training=True)
train_dataloader.build()
train_inputs = train_dataloader.get_one_batch()
#---validation
val_dataloader = DataLoader(train_config['validation_data_config'],
is_training=False)
val_dataloader.build()
val_inputs = val_dataloader.get_one_batch()
# Save configurations for future reference
save_cfgs(train_dir, model_config, train_config, track_config)
if train_config['lr_config'].get('lr_warmup', False):
warmup_epoch_num = 10
init_lr_ratio = 0.8
warmup_steps = warmup_epoch_num * int(
train_config['train_data_config']['num_examples_per_epoch']
) // train_config['train_data_config']['batch_size']
inc_per_step = (
1 - init_lr_ratio
) * train_config['lr_config']['initial_lr'] / warmup_steps
warmup_lr = train_config['lr_config'][
'initial_lr'] * init_lr_ratio + inc_per_step * tf.to_float(
global_step)
learning_rate = tf.cond(
tf.less(global_step,
warmup_steps), lambda: tf.identity(warmup_lr),
lambda: _configure_learning_rate(train_config, global_step -
warmup_steps))
else:
learning_rate = _configure_learning_rate(train_config, global_step)
optimizer = _configure_optimizer(train_config, learning_rate)
tf.summary.scalar('learning_rate', learning_rate)
# Set up the training ops
examplars, instances, gt_examplar_boxes, gt_instance_boxes = tf.split(train_inputs[0],num_gpus), \
tf.split(train_inputs[1],num_gpus), \
tf.split(train_inputs[2],num_gpus), \
tf.split(train_inputs[3],num_gpus)
if train_config['train_data_config'].get('time_decay', False):
time_intervals = tf.split(train_inputs[4], num_gpus)
tower_grads = []
with tf.variable_scope(tf.get_variable_scope()):
for i in range(num_gpus):
with tf.device('/gpu:%d' % i):
if train_config['train_data_config'].get(
'time_decay', False):
inputs = [
examplars[i], instances[i], gt_examplar_boxes[i],
gt_instance_boxes[i], time_intervals[i]
]
else:
inputs = [
examplars[i], instances[i], gt_examplar_boxes[i],
gt_instance_boxes[i]
]
model = tower_model(Model,
inputs,
model_config,
train_config,
mode='train')
# Reuse variables for the next tower.
tf.get_variable_scope().reuse_variables()
grads = optimizer.compute_gradients(model.total_loss)
tower_grads.append(grads)
grads = average_gradients(tower_grads)
#Clip gradient
gradients, tvars = zip(*grads)
clip_gradients, _ = tf.clip_by_global_norm(
gradients, train_config['clip_gradients'])
train_op = optimizer.apply_gradients(zip(clip_gradients, tvars),
global_step=global_step)
#Build validation model
with tf.device('/gpu:0'):
model_va = Model(model_config,
train_config,
mode='validation',
inputs=val_inputs)
model_va.build(reuse=True)
#Save Model setup
saver = tf.train.Saver(
tf.global_variables(),
max_to_keep=train_config['max_checkpoints_to_keep'])
summary_writer = tf.summary.FileWriter(train_dir, g)
summary_op = tf.summary.merge_all()
global_variables_init_op = tf.global_variables_initializer()
local_variables_init_op = tf.local_variables_initializer()
# Dynamically allocate GPU memory
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options,
allow_soft_placement=True)
#inter_op_parallelism_threads = 16, intra_op_parallelism_threads = 16, log_device_placement=True)
######Debug timeline
if Debug:
from tensorflow.python.client import timeline
run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
run_metadata = tf.RunMetadata()
######Debug timeline
sess = tf.Session(config=sess_config)
model_path = tf.train.latest_checkpoint(train_config['train_dir'])
if not model_path:
sess.run(global_variables_init_op)
sess.run(local_variables_init_op)
start_step = 0
if model_config['embed_config']['embedding_checkpoint_file']:
model.init_fn(sess)
elif model_config['finetuned_checkpoint_file']:
finetuned_checkpoint_file = tf.train.latest_checkpoint(
model_config['finetuned_checkpoint_file'])
logging.info('Restore from last checkpoint: {}'.format(
finetuned_checkpoint_file))
sess.run(local_variables_init_op)
sess.run(global_variables_init_op)
restore_op = tf.contrib.slim.assign_from_checkpoint_fn(
finetuned_checkpoint_file,
tf.global_variables(),
ignore_missing_vars=True)
restore_op(sess)
#reset global step saved in checkpoint
global_step_reset_op = global_step.assign(0)
sess.run(global_step_reset_op)
else:
logging.info('Restore from last checkpoint: {}'.format(model_path))
sess.run(local_variables_init_op)
sess.run(global_variables_init_op)
#saver.restore(sess, model_path)
restore_op = tf.contrib.slim.assign_from_checkpoint_fn(
model_path, tf.global_variables(), ignore_missing_vars=True)
restore_op(sess)
start_step = tf.train.global_step(sess, global_step.name) + 1
print_trainable(sess) #help function, can be disenable
g.finalize() # Finalize graph to avoid adding ops by mistake
# Training loop
data_config = train_config['train_data_config']
total_steps = int(data_config['epoch'] *
data_config['num_examples_per_epoch'] /
data_config['batch_size'])
logging.info('Train for {} steps'.format(total_steps))
for step in range(start_step, total_steps):
try:
start_time = time.time()
if Debug:
_, loss, batch_loss, current_lr = sess.run(
[
train_op, model.total_loss, model.batch_loss,
learning_rate
],
run_metadata=run_metadata,
options=run_options)
t1 = timeline.Timeline(run_metadata.step_stats)
ctf = t1.generate_chrome_trace_format()
with open('timeline.json', 'w') as f:
f.write(ctf)
else:
_, loss, batch_loss, current_lr = sess.run([
train_op, model.total_loss, model.batch_loss,
learning_rate
])
duration = time.time() - start_time
if step % 10 == 0:
examples_per_sec = data_config['batch_size'] / float(
duration)
time_remain = data_config['batch_size'] * (
total_steps - step) / examples_per_sec
current_epoch = (
step * data_config['batch_size']
) // data_config['num_examples_per_epoch'] + 1
m, s = divmod(time_remain, 60)
h, m = divmod(m, 60)
format_str = (
'%s: epoch %d-step %d,lr = %f, total loss = %.3f, batch loss = %.3f (%.1f examples/sec; %.3f '
'sec/batch; %dh:%02dm:%02ds remains)')
logging.info(
format_str %
(datetime.now(), current_epoch, step, current_lr, loss,
batch_loss, examples_per_sec, duration, h, m, s))
if step % 200 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
if step % train_config['save_model_every_n_step'] == 0 or (
step + 1) == total_steps:
checkpoint_path = osp.join(train_config['train_dir'],
'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
except KeyboardInterrupt:
checkpoint_path = osp.join(train_config['train_dir'],
'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
print("save model.ckpt-%d" % (step))
break
except:
print(traceback.format_exc())
print("Error found in current step, continue")
def main(checkpoint, input_files):
os.environ['CUDA_VISIBLE_DEVICES'] = auto_select_gpu()
model_config, _, track_config = load_cfgs(checkpoint)
track_config['log_level'] = 1
g = tf.Graph()
with g.as_default():
model = inference_wrapper.InferenceWrapper()
restore_fn = model.build_graph_from_config(model_config, track_config,
checkpoint)
g.finalize()
if not osp.isdir(track_config['log_dir']):
logging.info('Creating inference directory: %s',
track_config['log_dir'])
mkdir_p(track_config['log_dir'])
video_dirs = []
for file_pattern in input_files.split(","):
video_dirs.extend(glob(file_pattern))
logging.info("Running tracking on %d videos matching %s", len(video_dirs),
input_files)
gpu_options = tf.GPUOptions(allow_growth=True)
sess_config = tf.ConfigProto(gpu_options=gpu_options)
with tf.Session(graph=g, config=sess_config) as sess:
restore_fn(sess)
tracker = Tracker(model,
model_config=model_config,
track_config=track_config)
for video_dir in video_dirs:
if not osp.isdir(video_dir):
logging.warning(
'{} is not a directory, skipping...'.format(video_dir))
continue
video_name = osp.basename(video_dir)
video_log_dir = osp.join(track_config['log_dir'], video_name)
mkdir_p(video_log_dir)
filenames = sort_nicely(
glob(video_dir + '/img/*.jpg') +
glob(video_dir + '/img/*.png'))
first_line = open(video_dir + '/groundtruth_rect.txt').readline()
bb = [int(v) for v in first_line.strip().split(',')]
init_bb = Rectangle(bb[0] - 1, bb[1] - 1, bb[2],
bb[3]) # 0-index in python
trajectory = tracker.track(sess, init_bb, filenames, video_log_dir)
with open(osp.join(video_log_dir, 'track_rect.txt'), 'w') as f:
for region in trajectory:
rect_str = '{},{},{},{}\n'.format(region.x + 1,
region.y + 1,
region.width,
region.height)
f.write(rect_str)
with open(osp.join(video_log_dir, 'bboxes.json'), 'r') as f:
data = json.load(f)
final_output = {}
for i, fname in enumerate(data.keys()):
img = np.array(Image.open(fname).convert('RGB'))
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
#print(img,img.shape)
bboxes = data[fname]
bboxes = list(
map(
lambda x: list(
map(lambda y: float(y),
x.strip().split(','))), bboxes))
arr = []
for x, y, w, h in bboxes:
ymin, xmin, ymax, xmax = int(y), int(x), int(y +
h), int(x + w)
img = cv2.rectangle(img, (xmin, ymin), (xmax, ymax),
(0, 255, 0, 255), 2)
arr.append([ymin, xmin, ymax, xmax])
final_output[fname] = arr
name = osp.basename(fname)
name = osp.splitext(name)[0]
W, H, _ = img.shape
cv2.imshow("Pic", cv2.resize(img, (W // 2, H // 2)))
cv2.waitKey(0)
out_folder = osp.join(video_log_dir, "Outputs")
mkdir_p(out_folder)
cv2.imwrite(osp.join(out_folder, f"{name}_bbox.png"), img)
with open(osp.join(out_folder, "output.json"), "w") as f:
json.dump(final_output, f, indent=4)
cv2.destroyAllWindows()
