Image Processing Data Science Challenge

By |

Image Processing Data Science Challenge
Every Friday 1:00 pm to 5:00 pm  in Earl Lewis Room, Rackham Building
CSCAR is organizing a weekly group where participants can learn, sharpen, and
share their image processing, remote sensing, data analysis, and coding skills while
competing against professional and amateur data scientists. You should be familiar
with at least one of remote sensing, image processing, classification algorithms, and
programming. While working through the challenge, we will be using Python, R, and
Matlab. Final solutions have to be in Python or R since Kaggle only accepts solutions
in open source tools.
We are tackling image processing related challenges hosted at Kaggle (kaggle.com),
which is a leading platform to host data science challenges. It is currently running a
competition (https://www.kaggle.com/c/dstl-satellite-imagery-feature-detection)
where participants are asked to identify 10 different types of objects in very high
spatial resolution images collected by Worldview-3
(https://www.digitalglobe.com/) sensor. Some basic details of the data are as
follows.
• Wavebands
o Panchromatic: 450-800 nm
o 8 Multispectral: (red, red edge, coastal, blue, green, yellow, near-IR1 and
near-IR2) 400 nm – 1040 nm
o 8 SWIR: 1195 nm – 2365 nm
• Sensor Resolution (GSD) at Nadir :
o Panchromatic: 0.31m
o Multispectral: 1.24 m
o SWIR: Delivered at 7.5m
• Dynamic Range
o Panchromatic and multispectral : 11-bits per pixel
o SWIR : 14-bits per pixel
Depending on the interests in the group we will also include another image
processing challenge currently running at Kaggle sponsored by Nature Conservancy
– Fish Species Identification in digital camera images
(https://www.kaggle.com/c/the-nature-conservancy-fisheries-monitoring).
Send a mail to manishve@umich.edu if you want to participate or find out more
about it.

Geospatial analysis with Python

By |

This workshop will cover basic geospatial analysis in Python. Topics covered will include reading and writing various GIS file formats (shapefile, KML, geojson, csv), geocoding, common geometric operations like finding closest line to a point, point in polygon, spatial indexing and spatial joins etc. The workshop will focus solely on vector data (points, lines, polygons).  The will be mostly accomplished using the Python modules: fiona, shapely, rtree (but not arcpy).