else:
OD_MODEL_NAME=model
OD_MODEL_PATH='models/'+model+'/{}'
if optimized:
USE_OPTIMIZED=True
else:
USE_OPTIMIZED=False
if single_class:
NUM_CLASSES=1
else:
NUM_CLASSES=90
return TestConfig()
# Read sequentail Models or Gather all Models from models/
config = Config()
if config.SEQ_MODELS:
models = config.SEQ_MODELS
else:
for root, dirs, files in os.walk(MODELS_DIR):
if root.count(os.sep) - MODELS_DIR.count(os.sep) == 0:
for idx,model in enumerate(dirs):
models=[]
models.append(dirs)
models = np.squeeze(models)
models.sort()
print("> start testing following sequention of models: \n{}".format(models))
for mod in models:
print("> testing model: {}".format(mod))
# conditionals
def main():
model_type = 'dl'
input_type = 'image'
config = Config(model_type)
model = DeepLabModel(config).prepare_model(input_type)
model.run()
USE_OPTIMIZED = True
else:
USE_OPTIMIZED = False
if single_class:
NUM_CLASSES = 1
else:
NUM_CLASSES = 90
def __init__(self):
super(TestConfig, self).__init__(type)
return TestConfig()
# Read sequentail Models or Gather all Models from models/
config = Config('od')
if config.SEQ_MODELS:
model_names = config.SEQ_MODELS
else:
model_names = get_model_list(MODELS_DIR)
# Sequential testing
for model_name in model_names:
print("> testing model: {}".format(model_name))
# conditionals
optimized = False
single_class = False
# Test Model
if 'hands' in model_name or 'person' in model_name:
single_class = True
if 'deeplab' in model_name:
else:
masks = None
# reformat detection
num = int(num)
boxes = np.squeeze(boxes)
classes = np.squeeze(classes).astype(np.uint8)
scores = np.squeeze(scores)
# Visualization
vis = vis_detection(frame, boxes, classes,
scores, masks, category_index,
timer.get_fps(), config.VISUALIZE,
config.DET_INTERVAL, config.DET_TH,
config.MAX_FRAMES, None,
config.OD_MODEL_NAME + config._OPT)
if not vis:
break
cv2.destroyAllWindows()
timer.stop()
if __name__ == '__main__':
config = Config()
config.display()
model = Model('od', config.OD_MODEL_NAME, config.OD_MODEL_PATH,
config.LABEL_PATH, config.NUM_CLASSES, config.SPLIT_MODEL,
config.SSD_SHAPE).prepare_od_model()
detection(model, config)
def main():
config = Config()
model = Model('od', config.OD_MODEL_NAME, config.OD_MODEL_PATH,
config.LABEL_PATH, config.NUM_CLASSES, config.SPLIT_MODEL,
config.SSD_SHAPE).prepare_od_model()
detection(model, config)
def main():
model_type = 'od'
input_type = 'image'
config = Config(model_type)
model = ObjectDetectionModel(config).prepare_model(input_type)
model.run()
if 'detection_masks' in output_dict:
masks = output_dict['detection_masks'][0]
else:
masks = None
# reformat detection
num = int(num)
boxes = np.squeeze(boxes)
classes = np.squeeze(classes).astype(np.uint8)
scores = np.squeeze(scores)
# Visualization
vis = visualize_objectdetection(frame,boxes,classes,scores,masks,category_index,timer.get_frame(),
config.MAX_FRAMES,timer.get_fps(),config.PRINT_INTERVAL,config.PRINT_TH,
config.OD_DISPLAY_NAME,config.VISUALIZE,config.VIS_FPS,config.DISCO_MOE,config.ALPHA)
if not vis:
break
if config.SAVE_RESULT:
cv2.imwrite('{}/{}_{}.jpg'.format(config.RESULT_PATH,timer.get_frame(),config.OD_DISPLAY_NAME),frame)
cv2.destroyAllWindows()
timer.stop()
if __name__ == '__main__':
config = Config()
config.display()
model = Model('od', config.OD_MODEL_NAME, config.OD_MODEL_PATH, config.LABEL_PATH,
config.NUM_CLASSES, config.SPLIT_MODEL, config.SSD_SHAPE).prepare_od_model()
detection(model,config)
class TestConfig(Config):
LIMIT_IMAGES = 100
SPLIT_MODEL = False
WRITE_TIMELINE = True
if type is 'DL':
DL_MODEL_PATH = model + '/{}'
else:
OD_MODEL_PATH = model + '/{}'
if optimized:
USE_OPTIMIZED = True
return TestConfig()
# Read sequentail Models or Gather all Models from models/
CONFIG = Config()
if CONFIG.SEQ_MODELS:
models = CONFIG.SEQ_MODELS
else:
MODELS_DIR = os.path.join(os.getcwd(), 'models')
for root, dirs, files in os.walk(MODELS_DIR):
if root.count(os.sep) - MODELS_DIR.count(os.sep) == 0:
for idx, model in enumerate(dirs):
models = []
models.append(dirs)
models = np.squeeze(models)
models.sort()
print("> start testing following sequention of models: \n{}".format(models))
for mod in models:
print("> testing model: {}".format(mod))
MODEL_DIR = os.path.join(os.getcwd(), 'models', mod)
# -*- coding: utf-8 -*-
import tensorflow as tf
import yaml
import os
import sys
from rod.config import Config
## RUN THIS SCRIPT FROM THE ROOT DIR (NOT FROM /SCRIPTS)
MODEL_TYPE = 'od' # Change to 'OD' or 'DL'
NODE_NAMES = ['Momentum', 'Optimizer', 'BatchNorm', 'Loss']
NODE_OPS = ['Placeholder', 'Identity', 'CheckNumerics', 'BatchNorm']
## Don't Change ##
config = Config(MODEL_TYPE)
MODEL_PATH = config.MODEL_PATH
print("> exploring Model: {}".format(MODEL_PATH))
gf = tf.GraphDef()
gf.ParseFromString(open(MODEL_PATH, 'rb').read())
print('>Looking for node Names:')
for NAME in NODE_NAMES:
print(NAME)
print([n.name + '=>' + n.op for n in gf.node if NAME in n.name])
print('>Looking for node Ops:')
for OP in NODE_OPS:
print(OP)
