Apache Spark Accelerated Deep Learning Inference for Large Scale Satellite Image Analytics

Abstract

The shear volumes of data generated from earth observation and remote sensing technologies continue to make major impact; leaping key geospatial applications into the dual data and compute intensive era. As a consequence, this rapid advancement poses new computational and data processing challenges. We implement a novel remote sensing data flow (RESFlow) for advanced machine learning and computing with massive amounts of remotely sensed imagery. The core contribution is partitioning massive amount of data based on the spectral and semantic characteristics for distributed imagery analysis. RESFlow takes advantage of both a unified analytics engine for large-scale data processing and the availability of modern computing hardware to harness the acceleration of deep learning inference on expansive remote sensing imagery. The framework incorporates a strategy to optimize resource utilization across multiple executors assigned to a single worker. We showcase its deployment across computationally and data-intensive on pixel-level labeling workloads. The pipeline invokes deep learning inference at three stages; during deep feature extraction, deep metric mapping, and deep semantic segmentation. The tasks impose compute intensive and GPU resource sharing challenges motivating for a parallelized pipeline for all execution steps. By taking advantage of Apache Spark, Nvidia DGX1, and DGX2 computing platforms, we demonstrate unprecedented compute speed-ups for deep learning inference on pixel labeling workloads; processing 21,028 Terrabytes of imagery data and delivering an output maps at area rate of 5.245sq.km/sec, amounting to 453,168 sq.km/day -reducing a 28 day workload to 21 hours.

Multiobjective Task Scheduling for Energy-Efficient Cloud Implementation of Hyperspectral Image Classification

“…Finally, we can calculate the makespan and total energy consumption by Eqs. (3) and (5), respectively, to evaluate the fitness of this particular scheduling solution.…”

“…Due to the large data volume and high computational complexity of hyperspectral image applications, the processing of hyperspectral data, which generally involves computation-and data-intensive operations, naturally becomes a big data problem [4]. Motivated by the increasing demand for hyperspectral big data analytics, there emerges many research studies oriented toward the parallel implementation of hyperspectral image applications on cloud computing architectures [5]- [7]. The fundamental idea is to use the distributed file system to cope with the storage of hyperspectral big data, and utilize the distributed computing scheme to support the intensive computation during the processing flow.…”

A Big Data Framework for Satellite Images Processing using Apache Hadoop and RasterFrames: A Case Study of Surface Water Extraction in Phu Tho, Viet Nam

“…Recently, Dalton Dunga [21] proposed RESFlow framework that includes several modular components such as clustering and embedding, image-bucket assignment, image gallery, model gallery, accelerated inference, application space,and image analytic. The framework provides parallel computing using Spark.…”

Rolling Bearing Fault Diagnosis Method Based on Parallel QPSO-BPNN Under Spark-GPU Platform

“…Due to the many-core GPU has the advantages of highperformance, low-power, and low-cost, recently some work has been done to explore how to combine Spark and GPU to accelerate solving domain-specific applications, such as urban traffic vehicle recognition [28], magnetic resonance imaging [29], and remote sensing image processing [30], etc. The experimental results from [28], [29], [30] show that combining Spark and GPU can significantly improve the performance of these applications, but the implementations of them are complicated and it is difficult to port their implementation methods to other fields. The newly released Spark 3.0 already supports the accelerator-aware scheduling, allowing users to discover and request GPU computing resources at Executor, Driver, and Task levels, which simplifies the development of applications based on Spark and GPU.…”