keras_shufflenetv2's Introduction

keras_shufflenetv2

ShufflenetV2

Layer (type)	Output Shape	Param #
conv2d (Conv2D)	multiple	640
conv2d_1 (Conv2D)	multiple	1056
separable_conv2d (SeparableC	multiple	1344
max_pooling2d (MaxPooling2D)	multiple	0
dense (Dense)	multiple	6423040
dense_1 (Dense)	multiple	5130
batch_normalization (BatchNo	multiple	128
flatten (Flatten)	multiple	0
dropout (Dropout)	multiple	0

epoch = 200

acc = 94.57

CNN

Layer (type)	Output Shape	Param #
conv2d_2 (Conv2D)	multiple	320
conv2d_3 (Conv2D)	multiple	18496
max_pooling2d_1 (MaxPooling2	multiple	0
dense_2 (Dense)	multiple	6423040
dense_3 (Dense)	multiple	5130
flatten_1 (Flatten)	multiple	0
dropout_1 (Dropout)	multiple	0

epoch = 200

acc = 84.75

keras_shufflenetv2's People

Contributors

Stargazers

Watchers

keras_shufflenetv2's Issues

I upload only the edits I made

file name : train.py

def training_pipeline_ShuffleNetV2():

data_dir = 'insect_30'
image_size = 299

data_generator = ImageDataGenerator(rescale=1./255, validation_split=0.1)

train_generator = data_generator.flow_from_directory(
    str(data_dir),
    subset='training',
    target_size=(image_size, image_size),
    batch_size=model_config._BATCH_SIZE,
    class_mode='categorical',
    shuffle=True,
    seed=0)

validation_generator = data_generator.flow_from_directory(
    str(data_dir),
    subset='validation',
    target_size=(image_size, image_size),
    batch_size=model_config._BATCH_SIZE,
    class_mode='categorical',
    shuffle=True,
    seed=0)

model = ShuffleNetV2Model()  # your keras model here
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

model.fit_generator(
    train_generator,
    steps_per_epoch=?,
    epochs=model_config._EPOCHS,
    validation_data=validation_generator,
    validation_steps=?)

# display model config
model.summary()

if name == 'main':