val (train, test) = fashionMnist()
val inputs = Input(28, 28, 1)
val conv1 = Conv2D(32)(inputs)
val conv2 = Conv2D(64)(conv1)
val maxPool = MaxPool2D(poolSize = intArrayOf(1, 3, 3, 1), 
                        strides = intArrayOf(1, 3, 3, 1))(conv2)
val conv3 = Conv2D(64)(maxPool)
val conv4 = Conv2D(64)(conv3)
val add1 = Add()(conv4, maxPool)
val conv5 = Conv2D(64)(add1)
val conv6 = Conv2D(64)(conv5)
val add2 = Add()(conv6, add1)
val conv7 = Conv2D(64)(add2)
val globalAvgPool2D = GlobalAvgPool2D()(conv7)
val dense1 = Dense(256)(globalAvgPool2D)
val outputs = Dense(10, activation = Activations.Linear)(dense1)
val model = Functional.fromOutput(outputs)
model.use {
   it.compile(
       optimizer = Adam(),
       loss = Losses.SOFT_MAX_CROSS_ENTROPY_WITH_LOGITS,
       metric = Metrics.ACCURACY
   )
   it.summary()
   it.fit(dataset = train, epochs = 3, batchSize = 1000)
   val accuracy = it.evaluate(dataset = test, batchSize = 1000)
                    .metrics[Metrics.ACCURACY]
   println("Accuracy after: $accuracy")
}

// specify the model type to be loaded, ResNet50, for example
val loader =
   ModelZoo(commonModelDirectory = File("cache/pretrainedModels"), modelType = ModelType.ResNet_50)
// obtain the model configuration
val model = loader.loadModel() as Functional
// load class labels (from ImageNet dataset in ResNet50 case)
val imageNetClassLabels = loader.loadClassLabels()
// load weights if required (for Transfer Learning purposes)
val hdfFile = loader.loadWeights()

val model = resnet50Light(imageSize = 28, 
                          numberOfClasses = 10, 
                          numberOfChannels = 1, 
                          lastLayerActivation = Activations.Linear)

val preprocessing: Preprocessing = preprocess {
   transformImage {
       load {
           pathToData = imageDirectory
           imageShape = ImageShape(224, 224, 3)
           colorMode = ColorOrder.BGR
       }
       rotate {
           degrees = 30f
       }
       crop {
           left = 12
           right = 12
           top = 12
           bottom = 12
       }
       resize {
           outputWidth = 400
           outputHeight = 400
           interpolation = InterpolationType.NEAREST
       }
   }
   transformTensor {
       rescale {
           scalingCoefficient = 255f
       }
   }
}

repositories {
    mavenCentral()
}
dependencies {
    implementation 'org.jetbrains.kotlinx:kotlin-deeplearning-api:0.2.0'
}