ARM Compute Library v17.12 がリリースされています。
画像処理モジュールを中心に実装していたARM Compute Library。v17.06でMachine Learning関連を導入して、
今回リリースされた v17.12 でもその部分は強化していますね。
今回リリースされた v17.12 でもその部分は強化していますね。
引用 New graph nodes for NEON and OpenCL arm_compute::graph::BranchLayer arm_compute::graph::DepthConvertLayer arm_compute::graph::DepthwiseConvolutionLayer arm_compute::graph::DequantizationLayer arm_compute::graph::FlattenLayer arm_compute::graph::QuantizationLayer arm_compute::graph::ReshapeLayerが追加され、examplesディレクトリには、
graph_mobilenet.cpp graph_vgg16.cpp graph_vgg19.cppが追加されました。
また、GLES_COMPUTEを利用したカーネルも追加されました
引用 New OpenGL ES kernels / functions arm_compute::GCAbsoluteDifferenceKernel / arm_compute::GCAbsoluteDifference arm_compute::GCActivationLayerKernel / arm_compute::GCActivationLayer arm_compute::GCBatchNormalizationLayerKernel / arm_compute::GCBatchNormalizationLayer arm_compute::GCCol2ImKernel arm_compute::GCDepthConcatenateLayerKernel / arm_compute::GCDepthConcatenateLayer arm_compute::GCDirectConvolutionLayerKernel / arm_compute::GCDirectConvolutionLayer arm_compute::GCDropoutLayerKernel / arm_compute::GCDropoutLayer arm_compute::GCFillBorderKernel / arm_compute::GCFillBorder arm_compute::GCGEMMInterleave4x4Kernel / arm_compute::GCGEMMInterleave4x4 arm_compute::GCGEMMMatrixAccumulateBiasesKernel arm_compute::GCGEMMMatrixAdditionKernel / arm_compute::GCGEMMMatrixMultiplyKernel / arm_compute::GCGEMM arm_compute::GCGEMMTranspose1xWKernel / arm_compute::GCGEMMTranspose1xW arm_compute::GCIm2ColKernel arm_compute::GCNormalizationLayerKernel / arm_compute::GCNormalizationLayer arm_compute::GCPixelWiseMultiplicationKernel / arm_compute::GCPixelWiseMultiplication arm_compute::GCPoolingLayerKernel / arm_compute::GCPoolingLayer arm_compute::GCLogits1DMaxKernel / arm_compute::GCLogits1DShiftExpSumKernel arm_compute::GCLogits1DNormKernel / arm_compute::GCSoftmaxLayer arm_compute::GCTransposeKernel / arm_compute::GCTranspose