name: "Convolutional Neural Network"
input: "Images (1000x1x28x28)"
input_dim: 1000
input_dim: 3
input_dim: 28
input_dim: 28

layer {
	bottom: "Images (1000x1x28x28)"
	top: "Conv2d (1000x64x28x28)"
	name: "Conv2d (1000x64x28x28)"
	type: "Convolution"
	convolution_param {
		num_output: 64
		kernel_size: 5
		stride: 1
		pad: 2
	}
}

layer {
	bottom: "Conv2d (1000x64x28x28)"
	top: "Conv2d (1000x64x28x28)"
	name: "BatchNorm 1"
	type: "BatchNorm"
	batch_norm_param {
		use_global_stats: true
	}
}

layer {
	bottom: "Conv2d (1000x64x28x28)"
	top: "Conv2d (1000x64x28x28)"
	name: "ReLU 1"
	type: "ReLU"
}

layer {
	bottom: "Conv2d (1000x64x28x28)"
	top: "Conv2d (1000x64x28x28)"
	name: "MaxPool 1 (1000x64x14x14)"
	type: "Pooling"
	pooling_param {
		kernel_size: 3
		stride: 2
		pool: MAX
	}
}

layer {
	bottom: "Conv2d (1000x64x28x28)"
	top: "Conv2d (1000x128x14x14)"
	name: "Conv2d (1000x128x14x14)"
	type: "Convolution"
	convolution_param {
		num_output: 128
		kernel_size: 1
		pad: 0
		stride: 1
		bias_term: false
	}
}

layer {
	bottom: "Conv2d (1000x128x14x14)"
	top: "Conv2d (1000x128x14x14)"
	name: "BatchNorm 2"
	type: "BatchNorm"
	batch_norm_param {
		use_global_stats: true
	}
}

layer {
	bottom: "Conv2d (1000x128x14x14)"
	top: "Conv2d (1000x128x14x14)"
	name: "ReLU 2"
	type: "ReLU"
}

layer {
	bottom: "Conv2d (1000x128x14x14)"
	top: "Conv2d (1000x128x14x14)"
	name: "MaxPool 2 (1000x128x7x7)"
	type: "Pooling"
	pooling_param {
		kernel_size: 3
		stride: 2
		pool: MAX
	}
}


layer {
	bottom: "Conv2d (1000x128x14x14)"
	top: "Flatten (1000x6272)"
	name: "Flatten (1000x6272)"
	type: "InnerProduct"
	inner_product_param {
		num_output: 6272
	}
}

layer {
	bottom: "Flatten (1000x6272)"
	top: "Linear (1000x10)"
	name: "Linear (1000x10)"
	type: "InnerProduct"
	inner_product_param {
		num_output: 10
	}
}

layer {
	bottom: "Linear (1000x10)"
	top: "Softmax"
	name: "Softmax"
	type: "Softmax"
}