Classification accuracy assessment - Lab 7

How to perform a supervised landcover classification in Google Earth Engine. Landsat-8 image collected over Perth, Australia. Tutorial covers the collection of training and validation data, and deriving the confusion/error matrix.
Tutorial:
github.com/geospatialeco/GEAR...
Shaun R Levick
www.geospatialecology.com

[@giancarloalciaturi7635]

Very good tutorial.

[@ekoajimustiko4072]

Very good tutorial. Akhirnya saya bisa. Thanks

[@prof.biplabbiswas5576]

Super Liked your Videos. Thank you so much.

[@CessTenn]

Fantastic!

[@lugas4049]

Can I use this accuracy assessment for my ndvi classification?

[@abdulrashidhassan7649]

Dear Shaun, please can you provide us with a code to do a change detection, to see how the land cover change between two decade.

[@Alternatives_Universum]

Two questions arose in the course of this great video:
1. How can we overcome the limitation 5.000 pixels that can be used for classification and validation, respectively, for accumulating a larger training set/ validation set?
2. What do the first/ second/ third/ fourth number in the error matrix refer to?

[@MiguelHCBits]

1. You can use more than 5000 pixels to train o validate, but you can´t print to the console more than 5000 of those.
2. They refer to the classes (landcover), the main diagonal are the perfect match, in other words that diagonal match the true data (training and validation data) with the predict data.
I hope this is useful :)

[@nuranisaharfiana2775]

it's a nice video. But I have a question, how do you put the number of samples you use as training samples and validation samples? is it using a purposive sample method or what?

[@sermovies3217]

hey sorry to ask you, but is the method to collect training data like this also working if you have an image collection? because then the ".first()" command will not work and it gives you an error ("image.select(...).sampleRegions is not a function"). Would be nice if you could run such a classification for one week (so that the spectral values of each class are still similar) over a country. At least a nice try if it works on more than one image same time.

[@gallen7717]

Is it possible to perform a time series analysis on two supervised classifications of the same place at different time intervals to analyse to land use change?

[@amanlonare4319]

Did you get the answer to this?

[@user-ii2nu8ll7n]

Hello. May I ask a question？If I classify the cropland map by using some thresholds of NDVI value, how can I create a matrix or how can I assess the accuracy?

[@landflame]

create NDVI and add it into the bands,,, others are the same as the tutorial illustrates

[@Alternatives_Universum]

@@landflame How to add NDVI to the bands?