Single or a sequence of BigDL FeatureTransformers to be used. E.g. Resize(256, 256) -> CenterCrop(224, 224) -> ChannelNormalize(123, 117, 104, 1, 1, 1) -> MatToTensor()
Single or a sequence of BigDL FeatureTransformers to be used.
Single or a sequence of BigDL FeatureTransformers to be used. E.g. Resize(256, 256) -> CenterCrop(224, 224) -> ChannelNormalize(123, 117, 104, 1, 1, 1) -> MatToTensor()
Provides DataFrame-based API for image pre-processing and feature transformation. DLImageTransformer follows the Spark Transformer API pattern and can be used as one stage in Spark ML pipeline.
The input column can be either DLImageSchema.byteSchema or DLImageSchema.floatSchema. If using DLImageReader, the default format is DLImageSchema.byteSchema The output column is always DLImageSchema.floatSchema.