Ingesting Apache MXNet Gluon Deep Learning Results Via MQTT and Apache NiFi

Ingesting Apache MXNet Gluon Deep Learning Results Via MQTT and Apache NiFi

Summary:

Using a Pre-Trained Model in Apache MXNet Gluon Python 3 code to classify a webcam image captured and processed with OpenCV. In our Python script, we capture the image to disk and capture JSON metadata about the percentage, probabilities and device information. This JSON data is then sent via MQTT to a broker. Apache NiFi processes the JSON data.

Example Image

Source Code

• Schema: https://github.com/tspannhw/OpenSourceComputerVision/blob/master/gluon2.avsc
• Python Source: https://github.com/tspannhw/OpenSourceComputerVision/blob/master/nifigluon2.py
• Shell Script: https://github.com/tspannhw/OpenSourceComputerVision/blob/master/rungluon2.sh

SQL Table DDL

CREATE EXTERNAL TABLE IF NOT EXISTS gluon2 (top1pct STRING, top2pct STRING, top3pct STRING,
top4pct STRING, top5pct STRING, top1 STRING, top2 STRING, top3 STRING, top4 STRING,
top5 STRING, imgname STRING, host STRING, end STRING, te STRING, battery INT,
systemtime STRING, cpu DOUBLE, diskusage STRING, memory DOUBLE, id STRING)
STORED AS ORC
LOCATION '/gluon2'

Technologies: Python 3, Apache MXNet, Gluon, MQTT, Apache NiFi, OpenCV.

Based on http://gluon-crash-course.mxnet.io/predict.html

Apache NiFi Overview

'

Steps

1. ConsumeMQTT: Ingest MQTT data from gluon2 topic sent from Python
2. InferAvroSchema: One time grab the schema, then you can remove this processor.
3. RouteOnContent: Throw away errors
4. MergeRecord: Convert many JSON records into one large Apache AVRO file
5. ConvertAvroToORC: Convert that Apache AVRO File into an Apache ORC file
6. PutHDFS: Store the ApacheORC file in HDFS.

A side effect of the process is that is produces a SQL DDL to create a new table for this schema.

Table Example