def create_models(num_classes,
weights='imagenet',
multi_gpu=0,
snapshot_start=None):
# create "base" model (no NMS)
image = keras.layers.Input((None, None, 3))
# Keras recommends initialising a multi-gpu model on the CPU to ease weight sharing, and to prevent OOM errors.
# optionally wrap in a parallel model
if multi_gpu > 1:
with tf.device('/cpu:0'):
if snapshot_start == None:
model = ResNet50RetinaNet(image,
num_classes=num_classes,
weights=weights,
nms=False)
else:
model = snapshot_start
training_model = multi_gpu_model(model, gpus=multi_gpu)
else:
if snapshot_start == None:
model = ResNet50RetinaNet(image,
num_classes=num_classes,
weights=weights,
nms=False)
else:
model = snapshot_start
training_model = model
# append NMS for prediction only
classification = model.outputs[1]
detections = model.outputs[2]
boxes = keras.layers.Lambda(lambda x: x[:, :, :4])(detections)
detections = keras_retinanet.layers.NonMaximumSuppression(name='nms')(
[boxes, classification, detections])
prediction_model = keras.models.Model(inputs=model.inputs,
outputs=model.outputs[:2] +
[detections])
# Kai - check if we still need to compile if restarting from a snapshot
# compile model
training_model.compile(loss={
'regression': keras_retinanet.losses.smooth_l1(),
'classification': keras_retinanet.losses.focal()
},
optimizer=keras.optimizers.adam(lr=1e-5,
clipnorm=0.001))
return model, training_model, prediction_model
def create_models(num_classes, p):
# create "base" model (no NMS)
image = keras.layers.Input((None, None, 3))
if p['resnet'] == 101:
model = ResNet101RetinaNet(image, num_classes=num_classes, weights=p['weights'], nms=False)
elif p['resnet'] == 152:
model = ResNet152RetinaNet(image, num_classes=num_classes, weights=p['weights'], nms=False)
else: # 50
model = ResNet50RetinaNet(image, num_classes=num_classes, weights=p['weights'], nms=False)
training_model = model
# append NMS for prediction only
classification = model.outputs[1]
detections = model.outputs[2]
boxes = keras.layers.Lambda(lambda x: x[:, :, :4])(detections)
detections = layers.NonMaximumSuppression(name='nms')([boxes, classification, detections])
prediction_model = keras.models.Model(inputs=model.inputs, outputs=model.outputs[:2] + [detections])
# compile model
training_model.compile(
loss={
'regression' : losses.smooth_l1(),
'classification': losses.focal()
},
optimizer=keras.optimizers.adam(lr = p['learning-rate'], clipnorm=0.001),
metrics = []
)
return model, training_model, prediction_model
def create_models(num_classes):
# create "base" model (no NMS)
image = keras.layers.Input((None, None, 3))
model = ResNet50RetinaNet(image,
num_classes=num_classes,
weights='imagenet',
nms=False)
training_model = model
# append NMS for prediction only
classification = model.outputs[1]
detections = model.outputs[2]
boxes = keras.layers.Lambda(lambda x: x[:, :, :4])(detections)
detections = keras_retinanet.layers.NonMaximumSuppression(name='nms')(
[boxes, classification, detections])
prediction_model = keras.models.Model(inputs=model.inputs,
outputs=model.outputs[:2] +
[detections])
# compile model
training_model.compile(loss={
'regression': keras_retinanet.losses.smooth_l1(),
'classification': keras_retinanet.losses.focal()
},
optimizer=keras.optimizers.adam(lr=1e-5,
clipnorm=0.001))
return model, training_model, prediction_model
def create_models(num_classes, weights='imagenet', multi_gpu=0):
# create "base" model (no NMS)
image = keras.layers.Input((None, None, 3))
# Keras recommends initialising a multi-gpu model on the CPU to ease weight sharing, and to prevent OOM errors.
# optionally wrap in a parallel model
if args.multi_gpu > 1:
with tf.device('/cpu:0'):
model = ResNet50RetinaNet(image,
num_classes=num_classes,
weights=weights,
nms=False)
training_model = multi_gpu_model(model, gpus=args.multi_gpu)
else:
model = ResNet50RetinaNet(image,
num_classes=num_classes,
weights=weights,
nms=False)
training_model = model
# append NMS for prediction
detections = keras_retinanet.layers.NonMaximumSuppression(name='nms')(
model.outputs)
prediction_model = keras.models.Model(inputs=model.inputs,
outputs=model.outputs[:2] +
[detections])
# compile model
training_model.compile(loss={
'regression': keras_retinanet.losses.smooth_l1(),
'classification': keras_retinanet.losses.focal()
},
optimizer=keras.optimizers.adam(lr=1e-5,
clipnorm=0.001))
return model, training_model, prediction_model
def create_model(weights='imagenet'):
image = keras.layers.Input((None, None, 3))
return ResNet50RetinaNet(image, num_classes=20, weights=weights)