|
| 1 | +package org.janelia.render.client.spark.n5; |
| 2 | + |
| 3 | +import com.beust.jcommander.Parameter; |
| 4 | + |
| 5 | +import java.io.IOException; |
| 6 | +import java.io.Serializable; |
| 7 | +import java.nio.file.Paths; |
| 8 | +import java.util.Arrays; |
| 9 | +import java.util.List; |
| 10 | +import java.util.stream.Collectors; |
| 11 | + |
| 12 | +import org.apache.spark.SparkConf; |
| 13 | +import org.apache.spark.api.java.JavaSparkContext; |
| 14 | +import org.janelia.alignment.util.NeuroglancerAttributes; |
| 15 | +import org.janelia.render.client.ClientRunner; |
| 16 | +import org.janelia.render.client.parameter.CommandLineParameters; |
| 17 | +import org.janelia.saalfeldlab.n5.DatasetAttributes; |
| 18 | +import org.janelia.saalfeldlab.n5.N5Writer; |
| 19 | +import org.janelia.saalfeldlab.n5.spark.downsample.N5DownsamplerSpark; |
| 20 | +import org.janelia.saalfeldlab.n5.spark.supplier.N5WriterSupplier; |
| 21 | +import org.janelia.saalfeldlab.n5.universe.N5Factory; |
| 22 | +import org.slf4j.Logger; |
| 23 | +import org.slf4j.LoggerFactory; |
| 24 | + |
| 25 | +/** |
| 26 | + * Helper for downsampling a dataset. |
| 27 | + */ |
| 28 | +public class DownsampleHelper |
| 29 | + implements Serializable { |
| 30 | + |
| 31 | + private final String basePathOrStorageUrl; |
| 32 | + private final String sZeroDatasetPath; |
| 33 | + private final int[] downsampleFactors; |
| 34 | + private final int requiredSLevel; |
| 35 | + private final List<Double> stackResolutionValues; |
| 36 | + private final String stackResolutionUnit; |
| 37 | + private final List<Long> translatePixels; |
| 38 | + private final N5RetryUtil.RetryParameters retryParameters; |
| 39 | + |
| 40 | + /** |
| 41 | + * @param basePathOrStorageUrl base path or storage URL |
| 42 | + * e.g. gs://janelia-spark-test/hess_wafers_60_61_export or |
| 43 | + * /nrs/hess/data/hess_wafers_60_61/export/hess_wafers_60_61.n5 |
| 44 | + * |
| 45 | + * @param sZeroDatasetPath full-resolution dataset path |
| 46 | + * e.g. /flat/w61_serial_070_to_079/w61_s076_r00/raw_clahe/s0 |
| 47 | + * |
| 48 | + * @param downsampleFactors per-dimension factors applied at each downsampling step (e.g. 2,2,1). |
| 49 | + * |
| 50 | + * @param requiredSLevel the minimum s-level that must be produced (e.g. 9). |
| 51 | + * |
| 52 | + * @param retryParameters parameters controlling retry behavior on failure |
| 53 | + * (specify as null if you do not want retries performed). |
| 54 | + * |
| 55 | + * @throws IOException |
| 56 | + * if the sZeroDatasetPath does not end with '/s0'. |
| 57 | + */ |
| 58 | + public DownsampleHelper(final String basePathOrStorageUrl, |
| 59 | + final String sZeroDatasetPath, |
| 60 | + final int[] downsampleFactors, |
| 61 | + final int requiredSLevel, |
| 62 | + final List<Double> stackResolutionValues, |
| 63 | + final String stackResolutionUnit, |
| 64 | + final List<Long> translatePixels, |
| 65 | + final N5RetryUtil.RetryParameters retryParameters) |
| 66 | + throws IOException { |
| 67 | + |
| 68 | + this.basePathOrStorageUrl = basePathOrStorageUrl; |
| 69 | + |
| 70 | + this.sZeroDatasetPath = sZeroDatasetPath; |
| 71 | + if (! sZeroDatasetPath.endsWith("/s0")) { |
| 72 | + throw new IOException("sZeroDatasetPath must end with '/s0'"); |
| 73 | + } |
| 74 | + |
| 75 | + this.downsampleFactors = downsampleFactors; |
| 76 | + this.requiredSLevel = requiredSLevel; |
| 77 | + this.stackResolutionValues = stackResolutionValues; |
| 78 | + this.stackResolutionUnit = stackResolutionUnit; |
| 79 | + this.translatePixels = translatePixels; |
| 80 | + this.retryParameters = retryParameters; |
| 81 | + } |
| 82 | + |
| 83 | + /** |
| 84 | + * Downsamples the N5 dataset iteratively, creating s1, s2, ... and writing neuroglancer attributes. |
| 85 | + * Downsampling will stop when the sN result contains a single block is greater than or equal to the requiredSLevel. |
| 86 | + * |
| 87 | + * @param sparkContext the Spark context used for distributed processing. |
| 88 | + * |
| 89 | + * @throws IOException |
| 90 | + * if an N5 read or write operation fails. |
| 91 | + */ |
| 92 | + public void run(final JavaSparkContext sparkContext) |
| 93 | + throws IOException { |
| 94 | + |
| 95 | + LOG.info("run: entry, basePathOrStorageUrl={}, sZeroDatasetPath={}, downsampleFactors={}, requiredSLevel={}, retryParameters={}", |
| 96 | + basePathOrStorageUrl, sZeroDatasetPath, Arrays.toString(downsampleFactors), requiredSLevel, retryParameters); |
| 97 | + |
| 98 | + final N5WriterSupplier n5Supplier = () -> |
| 99 | + new N5Factory().openWriter(N5Factory.StorageFormat.N5, basePathOrStorageUrl); |
| 100 | + |
| 101 | + final N5Writer n5 = n5Supplier.get(); |
| 102 | + final DatasetAttributes fullScaleAttributes = n5.getDatasetAttributes(sZeroDatasetPath); |
| 103 | + final long[] dimensions = fullScaleAttributes.getDimensions(); |
| 104 | + final int numberOfDimensions = dimensions.length; |
| 105 | + final int[] outputBlockSize = fullScaleAttributes.getBlockSize(); |
| 106 | + final String outputGroupPath = sZeroDatasetPath.substring(0, sZeroDatasetPath.lastIndexOf('/')); |
| 107 | + |
| 108 | + int numberOfDownsampledDatasets = 0; |
| 109 | + long downsampledBlockCount = 2; |
| 110 | + for (int scale = 1; (downsampledBlockCount > 1) || scale <= requiredSLevel; scale++) { |
| 111 | + |
| 112 | + final String fromDataset = scale == 1 ? sZeroDatasetPath : outputGroupPath + "/s" + (scale - 1); |
| 113 | + final String toDataset = outputGroupPath + "/s" + scale; |
| 114 | + |
| 115 | + final int[] scaleFactors = new int[numberOfDimensions]; |
| 116 | + for (int d = 0; d < numberOfDimensions; d++) { |
| 117 | + scaleFactors[d] = (int) Math.round(Math.pow(downsampleFactors[d], scale)); |
| 118 | + } |
| 119 | + |
| 120 | + long blockCount = 1; |
| 121 | + final long[] downsampledDimensions = new long[numberOfDimensions]; |
| 122 | + for (int d = 0; d < numberOfDimensions; d++) { |
| 123 | + downsampledDimensions[d] = dimensions[d] / scaleFactors[d]; |
| 124 | + final long blocksInDim = (downsampledDimensions[d] + outputBlockSize[d] - 1) / outputBlockSize[d]; |
| 125 | + blockCount *= blocksInDim; |
| 126 | + } |
| 127 | + downsampledBlockCount = blockCount; |
| 128 | + |
| 129 | + if (n5.datasetExists(toDataset)) { |
| 130 | + |
| 131 | + final DatasetAttributes toDatasetAttributes = n5.getDatasetAttributes(toDataset); |
| 132 | + final long[] toDatasetDimensions = toDatasetAttributes.getDimensions(); |
| 133 | + for (int d = 0; d < numberOfDimensions; d++) { |
| 134 | + if (toDatasetDimensions[d] != downsampledDimensions[d]) { |
| 135 | + throw new IOException( |
| 136 | + "existing dataset " + toDataset + " has " + toDatasetDimensions[d] + |
| 137 | + " pixels in axis " + d + " instead of " + downsampledDimensions[d] + |
| 138 | + " pixels (based on downsampleFactor " + downsampleFactors[d] + ")"); |
| 139 | + } |
| 140 | + } |
| 141 | + |
| 142 | + LOG.info("run: skipping s{} because {} already exists", scale, toDataset); |
| 143 | + numberOfDownsampledDatasets++; |
| 144 | + |
| 145 | + continue; |
| 146 | + } |
| 147 | + |
| 148 | + final String operationDescription = "downsample " + fromDataset + " to " + toDataset; |
| 149 | + LOG.info("run: {} with {} downsampled block(s)", operationDescription, downsampledBlockCount); |
| 150 | + |
| 151 | + if (retryParameters == null) { |
| 152 | + N5DownsamplerSpark.downsample(sparkContext, |
| 153 | + n5Supplier, |
| 154 | + fromDataset, |
| 155 | + toDataset, |
| 156 | + downsampleFactors, |
| 157 | + null); |
| 158 | + } else { |
| 159 | + |
| 160 | + try { |
| 161 | + final RetryStats retryStats = N5RetryUtil.executeWithRetryVoid( |
| 162 | + () -> N5DownsamplerSpark.downsample(sparkContext, |
| 163 | + n5Supplier, |
| 164 | + fromDataset, |
| 165 | + toDataset, |
| 166 | + downsampleFactors), |
| 167 | + retryParameters, |
| 168 | + operationDescription); |
| 169 | + |
| 170 | + LOG.info("run: {}", retryStats); |
| 171 | + |
| 172 | + } catch (final Exception e) { |
| 173 | + throw new IOException(e); |
| 174 | + } |
| 175 | + |
| 176 | + } |
| 177 | + |
| 178 | + numberOfDownsampledDatasets++; |
| 179 | + } |
| 180 | + |
| 181 | + // save additional parameters so that n5 can be viewed in neuroglancer |
| 182 | + final NeuroglancerAttributes ngAttributes = |
| 183 | + new NeuroglancerAttributes(stackResolutionValues, |
| 184 | + stackResolutionUnit, |
| 185 | + numberOfDownsampledDatasets, |
| 186 | + downsampleFactors, |
| 187 | + translatePixels, |
| 188 | + NeuroglancerAttributes.NumpyContiguousOrdering.FORTRAN); |
| 189 | + |
| 190 | + ngAttributes.write(n5Supplier.get(), Paths.get(sZeroDatasetPath)); |
| 191 | + |
| 192 | + LOG.info("run: exit, generated {} downsampled datasets for {}", numberOfDownsampledDatasets, outputGroupPath); |
| 193 | + } |
| 194 | + |
| 195 | + public static class Parameters |
| 196 | + extends CommandLineParameters { |
| 197 | + |
| 198 | + @Parameter( |
| 199 | + names = "--basePathOrStorageUrl", |
| 200 | + description = "Base path or storage URL, e.g. gs://janelia-spark-test/hess_wafers_60_61_export or " + |
| 201 | + "/nrs/hess/data/hess_wafers_60_61/export/hess_wafers_60_61.n5", |
| 202 | + required = true) |
| 203 | + public String basePathOrStorageUrl; |
| 204 | + |
| 205 | + @Parameter( |
| 206 | + names = "--fullResolutionDataset", |
| 207 | + description = "Full-resolution dataset path, e.g. /flat/w61_serial_070_to_079/w61_s076_r00/raw_clahe/s0", |
| 208 | + required = true) |
| 209 | + public String fullResolutionDataset; |
| 210 | + |
| 211 | + @Parameter( |
| 212 | + names = "--factors", |
| 213 | + description = "Scale pyramid with given factors, e.g. 2,2,1", |
| 214 | + required = true) |
| 215 | + public String factors; |
| 216 | + |
| 217 | + @Parameter( |
| 218 | + names = "--requiredSLevel", |
| 219 | + description = "The minimum s-level that must be produced, e.g. 9") |
| 220 | + public int requiredSLevel = 0; |
| 221 | + |
| 222 | + @Parameter( |
| 223 | + names = "--stackResolution", |
| 224 | + description = "Resolution of the full scale x, y, and z axis pixels, e.g. 8,8,8", |
| 225 | + required = true) |
| 226 | + public String stackResolution; |
| 227 | + |
| 228 | + @Parameter( |
| 229 | + names = "--stackResolutionUnit", |
| 230 | + description = "Unit description for stack resolution values, e.g. nm, um, ...") |
| 231 | + public String stackResolutionUnit = "nm"; |
| 232 | + |
| 233 | + @Parameter( |
| 234 | + names = "--translate", |
| 235 | + description = "Translation pixels for the full scale x, y, and z axis, e.g. 100,-77,1. " + |
| 236 | + "Omit if translation is not needed.") |
| 237 | + public String translate; |
| 238 | + |
| 239 | + public int[] getDownsampleFactors() { |
| 240 | + return Util.parseCSIntArray(factors); |
| 241 | + } |
| 242 | + |
| 243 | + public List<Double> getStackResolutionValues() { |
| 244 | + return Arrays.stream(Util.parseCSIntArray(stackResolution)).asDoubleStream() |
| 245 | + .boxed().collect(Collectors.toList()); |
| 246 | + |
| 247 | + } |
| 248 | + |
| 249 | + public List<Long> getTranslatePixels() { |
| 250 | + return Arrays.stream(Util.parseCSIntArray(translate)).asLongStream() |
| 251 | + .boxed().collect(Collectors.toList()); |
| 252 | + } |
| 253 | + } |
| 254 | + |
| 255 | + public static void main(final String[] args) throws Exception { |
| 256 | + final ClientRunner clientRunner = new ClientRunner(args) { |
| 257 | + @Override |
| 258 | + public void runClient(final String[] args) |
| 259 | + throws Exception { |
| 260 | + |
| 261 | + final Parameters parameters = new Parameters(); |
| 262 | + parameters.parse(args); |
| 263 | + |
| 264 | + LOG.info("runClient: entry, parameters={}", parameters); |
| 265 | + |
| 266 | + final DownsampleHelper helper = new DownsampleHelper(parameters.basePathOrStorageUrl, |
| 267 | + parameters.fullResolutionDataset, |
| 268 | + parameters.getDownsampleFactors(), |
| 269 | + parameters.requiredSLevel, |
| 270 | + parameters.getStackResolutionValues(), |
| 271 | + parameters.stackResolutionUnit, |
| 272 | + parameters.getTranslatePixels(), |
| 273 | + new N5RetryUtil.RetryParameters()); |
| 274 | + |
| 275 | + final SparkConf conf = new SparkConf().setAppName("DownsampleHelper"); |
| 276 | + final JavaSparkContext sparkContext = new JavaSparkContext(conf); |
| 277 | + helper.run(sparkContext); |
| 278 | + sparkContext.close(); |
| 279 | + } |
| 280 | + }; |
| 281 | + clientRunner.run(); |
| 282 | + } |
| 283 | + |
| 284 | + private static final Logger LOG = LoggerFactory.getLogger(DownsampleHelper.class); |
| 285 | + |
| 286 | +} |
0 commit comments