def main(hp, model_id, save_base_dir, data_dir, train_epochs, total_epochs, epoch_index): # Xinyi modified
tf.logging.set_verbosity(tf.logging.ERROR)
model_dir = save_base_dir + str(model_id)
for name in list(flags.FLAGS):
delattr(flags.FLAGS, name)
define_cifar_flags(hp, model_id, model_dir, data_dir, train_epochs, total_epochs, epoch_index)
absl_app.parse_flags_with_usage(sys.argv)
return start(0)
def parse_flags(argv=None):
"""Reset flags and reparse. Currently only used in testing."""
flags.FLAGS.unparse_flags()
absl_app.parse_flags_with_usage(argv or sys.argv)
"Uses --dataset_path and --feature_spec"
"Overrides synthetic dataset dimension flags, other than the number of batches")
flags.DEFINE_integer('synthetic_dataset_train_batches', default=64008,
help='Number of training batches in the synthetic dataset')
flags.DEFINE_integer('synthetic_dataset_valid_batches', default=1350,
help='Number of validation batches in the synthetic dataset')
flags.DEFINE_list('synthetic_dataset_cardinalities', default=26*[1000],
help='Number of categories for each embedding table of the synthetic dataset')
flags.DEFINE_integer('synthetic_dataset_num_numerical_features', default=13,
help='Number of numerical features of the synthetic dataset')
define_command_line_flags()
FLAGS = flags.FLAGS
app.define_help_flags()
app.parse_flags_with_usage(sys.argv)
if FLAGS.xla:
if FLAGS.cpu:
os.environ['TF_XLA_FLAGS'] = '--tf_xla_auto_jit=fusible --tf_xla_cpu_global_jit'
else:
os.environ['TF_XLA_FLAGS'] = '--tf_xla_auto_jit=fusible'
import time
from lr_scheduler import LearningRateScheduler
import tensorflow as tf
import tensorflow_addons as tfa
import numpy as np
from utils import IterTimer, init_logging, dist_print
from dataloader import create_input_pipelines
def _flags_parser(argv):
argv = _benchmark.Initialize(argv)
return app.parse_flags_with_usage(argv)
def parse_flags(argv=None): """Reset flags and reparse. Currently only used in testing.""" flags.FLAGS.unparse_flags() absl_app.parse_flags_with_usage(argv or sys.argv)
def flags_parser(args):
# Plumbs the flags defined in this file to the main module, mostly for the
# console script wrapper tb-gcp-uploader.
for flag in set(flags.FLAGS.get_key_flags_for_module(__name__)):
flags.FLAGS.register_flag_by_module(args[0], flag)
return app.parse_flags_with_usage(args)
import tensorflow as tf
import numpy as np
from PIL import Image
from absl import flags, app
from yolov3_tf2.models import (YoloV3, YoloV3Tiny)
from yolov3_tf2.utils import draw_outputs, load_darknet_weights
from yolov3_tf2.dataset import transform_images
app.parse_flags_with_usage(['yolo_iou_threshold'])
def get_model(path):
yolo = YoloV3(classes=80)
yolo.load_weights(path)
return yolo
def find_cars(image, yolo):
#img_arr = np.array(image_orig)
if type(image) == list:
images = np.array([np.array(i) for i in image])
image = transform_images(images, 416)
else:
image = tf.expand_dims(np.array(image), axis=0)
image = transform_images(image, 416)
boxes, scores, classes, nums = yolo.predict(image, steps=1)
filtered = []
