Farm Pin Crop Detection Challenge
$10,000 USD
Classify fields in South Africa by crop type using Sentinel-2 satellite imagery
4 March–1 September 2019 23:59
612 data scientists enrolled, 68 on the leaderboard
Ideas and suggestions on working with dataset (imaging Software, mapping, general info)
published 1 May 2019, 11:34

NO SOLUTIONS HERE, only general info and dataset-related questions

edited 9 minutes later

For the start: images are in SENTINEL SAFE format:, you can open in, for example, with standard toolbox, or in QGIS software and use some tutorial like this one: If someone found better/easier way to visualize the images, please reply.

I'm currently working through the data with geopandas on the shape file and about to start poking at the Sentinel files with gdal module. I also had a look at the images and shapefiles in QGIS to get a sense of what it all looks like before I start trying to script the feature extraction. I also saw (as other people have posted) that a bunch of the form polygons lie outside of the SENTINEL images provided. I'm fairly sure that the overlay is correct though as the farms clearly sit along the riverbanks the entire way.

@Allank I'm also trying to get started using geopandas and rasterio.

However, I'm 100% noob to GIS data, so quite a steep learning curve here. My current question / task is using the polygons in the .shp files to mask the raster (.jp2) data.

Any ideas of how to go about doing that?

@allank, seems you're right about that part of dataset out of the map. Fields clearly follows the river, so overlay seems to be ok. It looks like we've got only part of the images (there should be more images to the right of the map), but given dataset info covers both maps. @ RenierBotha you can start with QGIS to just look through the data, because it has plenty of tools to give you more suggestions and it is very easy to learn. Then geopanda and gdal (@Allank already mentioned that). Anyway, this competition is a great way to learn new tricks!

Hi @RenierBotha I am pretty new to GIS, have you managed to read the .SAFE files in python? if you can share how did you do that? I tried to check .jp2 data but I am not sure if they are the only files we should care about?