forked from intel-analytics/BigDL-2.x
-
Notifications
You must be signed in to change notification settings - Fork 2
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
add maskrcnn inference example (intel-analytics#2944)
* add maskrcnn inference example * meet pr comments * add model download url
- Loading branch information
1 parent
bac5997
commit 811f063
Showing
11 changed files
with
294 additions
and
92 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
71 changes: 71 additions & 0 deletions
71
spark/dl/src/main/scala/com/intel/analytics/bigdl/models/maskrcnn/README.md
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,71 @@ | ||
| # MaskRCNN | ||
| This example demonstrates how to use BigDL to evaluate the [MaskRCNN](https://arxiv.org/abs/1703.06870) architecture on COCO data | ||
|
|
||
| ## Prepare the data | ||
| * You can download [COCO dataset](<http://cocodataset.org/>) firstly. | ||
| Extract the dataset and get images and annotations like (use **coco_2017_val** as example): | ||
| ``` | ||
| coco | ||
| |_ coco_val2017 | ||
| | |_ <im-1-name>.jpg | ||
| | |_ ... | ||
| | |_ <im-N-name>.jpg | ||
| |_ annotations | ||
| |_ instances_train2017.json | ||
| |_ ... | ||
| ``` | ||
|
|
||
| * Generate the hadoop sequence files for COCO dataset | ||
| The following command will transform the images and annotations into hadoop sequence files. | ||
| ```bash | ||
| java -cp com.intel.analytics.bigdl.models.utils.COCOSeqFileGenerator bigdl-VERSION-jar-with-dependencies.jar -f ./coco/coco_val2017 -m ./coco/annotations/instances_val2017.json -p 4 -o ./coco/output | ||
| ``` | ||
| In the above commands: | ||
| -f: the COCO image files location | ||
| -m: the annotation json file location | ||
| -o: generated seq files location | ||
| -p: number of parallel | ||
|
|
||
| ## Data Processing | ||
| Input data are transformed by several pipeline classes, such as ScaleResize, ChannelNormalize, ImageFeatureToBatch, etc. | ||
|
|
||
| ## Model | ||
| You can download **preTrain-MaskRCNN model** for BigDL by running | ||
| ```bash | ||
| wget https://bigdlmodels.s3-us-west-2.amazonaws.com/segmentation/bigdl_mask-rcnn_COCO_0.10.0.model | ||
| ``` | ||
| This MaskRCNN model refers to [facebookresearch/maskrcnn-benchmark](https://github.com/facebookresearch/maskrcnn-benchmark), and the model backbone is **R-50-FPN**. | ||
|
|
||
| ## Test the Model | ||
| * Spark standalone, example command | ||
| ```bash | ||
| spark-submit \ | ||
| --master spark://xxx.xxx.xxx.xxx:xxxx \ | ||
| --executor-cores cores_per_executor \ | ||
| --total-executor-cores total_cores_for_the_job \ | ||
| --driver-class-path dist/lib/bigdl-VERSION-jar-with-dependencies.jar \ | ||
| --class com.intel.analytics.bigdl.models.maskrcnn.Test \ | ||
| dist/lib/bigdl-VERSION-jar-with-dependencies.jar \ | ||
| --batchSize batch_size \ | ||
| -f hdfs://.../coco/val \ | ||
| --model modelPath | ||
| ``` | ||
| * Spark yarn client mode, example command | ||
| ```bash | ||
| spark-submit \ | ||
| --master yarn \ | ||
| --deploy-mode client \ | ||
| --executor-cores cores_per_executor \ | ||
| --num-executors executors_number \ | ||
| --driver-class-path dist/lib/bigdl-VERSION-jar-with-dependencies.jar \ | ||
| --class com.intel.analytics.bigdl.models.inception.Test \ | ||
| dist/lib/bigdl-VERSION-jar-with-dependencies.jar \ | ||
| --batchSize batch_size \ | ||
| -f hdfs://.../coco/val \ | ||
| --model modelPath | ||
| ``` | ||
| In the above command | ||
| * -f: where you put your COCO data, it should be a hdfs folder | ||
| * --model: the model snapshot file | ||
| * --batchSize: The mini-batch size. It is expected that the mini-batch size is a multiple of node_number * core_number. | ||
| * --partitionNum: the partition number, default is node_number * core_number. |
89 changes: 89 additions & 0 deletions
89
spark/dl/src/main/scala/com/intel/analytics/bigdl/models/maskrcnn/Test.scala
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,89 @@ | ||
| /* | ||
| * Copyright 2016 The BigDL Authors. | ||
| * | ||
| * Licensed under the Apache License, Version 2.0 (the "License"); | ||
| * you may not use this file except in compliance with the License. | ||
| * You may obtain a copy of the License at | ||
| * | ||
| * http://www.apache.org/licenses/LICENSE-2.0 | ||
| * | ||
| * Unless required by applicable law or agreed to in writing, software | ||
| * distributed under the License is distributed on an "AS IS" BASIS, | ||
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| * See the License for the specific language governing permissions and | ||
| * limitations under the License. | ||
| */ | ||
|
|
||
| package com.intel.analytics.bigdl.models.maskrcnn | ||
|
|
||
| import com.intel.analytics.bigdl.models.resnet.Utils.{TestParams, _} | ||
| import com.intel.analytics.bigdl.transform.vision.image._ | ||
| import com.intel.analytics.bigdl.transform.vision.image.augmentation._ | ||
| import com.intel.analytics.bigdl.utils.{Engine, T} | ||
| import scopt.OptionParser | ||
| import com.intel.analytics.bigdl.dataset.{DataSet, MiniBatch, segmentation} | ||
| import com.intel.analytics.bigdl.nn.Module | ||
| import com.intel.analytics.bigdl.optim.MeanAveragePrecision | ||
| import org.apache.spark.{SparkContext, rdd} | ||
| object Test { | ||
| case class TestParams( | ||
| folder: String = "./", | ||
| model: String = "", | ||
| batchSize: Int = 2, | ||
| partitionNum: Int = -1 | ||
| ) | ||
|
|
||
| val testParser = new OptionParser[TestParams]("BigDL Mask-RCNN on COCO Test Example") { | ||
| opt[String]('f', "folder") | ||
| .text("the location of COCO dataset") | ||
| .action((x, c) => c.copy(folder = x)) | ||
|
|
||
| opt[String]('m', "model") | ||
| .text("the location of model snapshot") | ||
| .action((x, c) => c.copy(model = x)) | ||
|
|
||
| opt[Int]('b', "batchSize") | ||
| .text("total batch size") | ||
| .action((x, c) => c.copy(batchSize = x)) | ||
|
|
||
| opt[Int]('p', "partitionNum") | ||
| .text("partition number") | ||
| .action((x, c) => c.copy(partitionNum = x)) | ||
| } | ||
|
|
||
| def main(args: Array[String]): Unit = { | ||
| testParser.parse(args, TestParams()).foreach { param => { | ||
| val conf = Engine.createSparkConf().setAppName("Test MaskRCNN on COCO") | ||
| .set("spark.akka.frameSize", 64.toString) | ||
| .set("spark.task.maxFailures", "1") | ||
| val sc = new SparkContext(conf) | ||
| Engine.init | ||
|
|
||
| val partitionNum = if (param.partitionNum > 0) param.partitionNum | ||
| else Engine.nodeNumber() * Engine.coreNumber() | ||
|
|
||
| val rddData = DataSet.SeqFileFolder.filesToRoiImageFrame(param.folder, sc, Some(partitionNum)) | ||
| .toDistributed().data(train = false) | ||
|
|
||
| val transformer = MTImageFeatureToBatchWithResize( | ||
| sizeDivisible = 32, | ||
| batchSize = param.batchSize / Engine.nodeNumber(), | ||
| transformer = | ||
| PixelBytesToMat() -> | ||
| ScaleResize(minSize = 800, maxSize = 1333) -> | ||
| ChannelNormalize(122.7717f, 115.9465f, 102.9801f) -> | ||
| MatToTensor[Float](), | ||
| toRGB = false | ||
| ) | ||
| val evaluationSet = transformer(rddData) | ||
|
|
||
| val model = Module.loadModule[Float](param.model) | ||
|
|
||
| val result = model.evaluate(evaluationSet, | ||
| Array(MeanAveragePrecision.cocoBBox(81), MeanAveragePrecision.cocoSegmentation(81))) | ||
| result.foreach(r => println(s"${r._2} is ${r._1}")) | ||
|
|
||
| sc.stop() | ||
| }} | ||
| } | ||
| } |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.