DEEP LEARNING in QGIS: Image Segmentation (Aerial and Satellite) with the DEEPNESS Plugin

[@geospatialschool]

FREE GIS Courses: geospatialschool.com

[@hectorpatino9205]

If you're encountering the exception "unsupported layer units," the issue may be due to the coordinate system in use. After reviewing the logs, I found that the problem originates in the processing_utils.py file. Try reprojecting the source layer to a coordinate system other than Long/Lat.

[@lc4468]

Daaaaaaaang. It worked surprisingly well on .45m imagery. Thanks for sharing!

[@geospatialschool]

Great to hear!

[@GEORGE.M.M]

This is a massive time saver, thank you!

[@geospatialschool]

You're welcome. Glad it's helpful!

[@AliciaLangton-u5l]

Thank you for this. Any chance you have had success installing the plugin on a Mac? I'm having to use it on my windows partition since the dependencies would not install automatically.

[@Testtest-eg9sq]

hello, I experienced a Traceback (most recent call last) when running the deepness plugin. the description is in convert_meters_to_rlayer_units raise Exception("Unsupported layer units") can you help me with the solution?

[@tiaraysuff]

Dear sir, I'm still new about deep learning and I have a question from this video, can deep learning such as convolutional neural networks (CNN) be operated using this plugin?

[@jupita3000]

What if I only wanted a 2 class output? Ex. pervious and impervious. How would I specify that? Is it possible with this plugin?

[@axeLeonhardt]

Are there any pre-trained models available for high-resolution drone images, or does it require training? thanks for sharing

[@geospatialschool]

There aren't any with this plug-in. You you may be able to find some elsewhere

[@agrengs0]

Thanks for the video. I imported a high res tiff and getting an error "Exception: unsupported layer units"

[@mohamedelgamal6333]

Dear Sir, Many thanks for this greate video. I followed the stepa you mentioned and I installed the deepness plugin successfully and test it then when I triedd to apply it on a rater image it gives python error and a message " This plugin is incompatible with this version of QGIS Plugin designed for QGIS 3.22 - 3.99" despite I am using QGIS 3.18. Any help will be appreciated.

[@mohamedelgamal6333]

Many thanks Sir, I figured out that it seems I need to reinstall some of the other python package couple of times and it is now working. Thank you again

[@geospatialschool]

You need to update we qgis so that your version is at least 3.22

[@sridharan19]

Sir can we map demolition and construction waste using this?

[@kyawnaingwin8300]

To my understanding, you have to train your detection model first with sample images with lable boxes of your object of interest. Then you get a model and use on QGIS here.

[@taufiqhidayat628]

can i get the error value?

[@coolsamuel4354]

Can the first model work on images from Google? Thanks for the video

[@geospatialschool]

Maybe. You'll have to give it a try and perform some accuracy analysis. As with any model, it's going to be wrong sometimes. You'll have to determine if it will work for you specific application.

[@coolsamuel4354]

@@geospatialschool thanks I tried it but I guess it was designed for Poland and a bit of Europe. The result accuracy was very poor

[@desireespindler3958]

Thank you for this great video. I have a problem, when I wannt to load the model deeplabv3_landcover_4c.onnx on Deepnes, I get this error "Error! Failed to load the model! Model may be not usable." Why could it be? From already thank you very much

[@geospatialschool]

To be honest, I'm not sure. You'll have to look into the deepness documentation and ask their team

[@nadezhda_belyaeva_]

I had the same error when using Qgis version 3.34, and with version Qgis 3.28 model loaded without any problems

[@dauletkairatov4620]

I’ve got same problem, if u have a chance can u gimme some tips how to resolve this problem

[@lorenzoconti7372]

there are any object based classification model for Landsat ?

[@geospatialschool]

Object-based classification is usually implemented with high res data where objects (like trees or buildings) are composed of many pixels. Landsat is coarse enough that pixel-based methods are generally used, though I'm sure there are some object based approaches if you look around

[@amritanshuraj1]

thank you, this is very helpful

[@geospatialschool]

You're welcome!

[@emelka5409]

can i get your contact?? i want to discuss more with you

[@BongaLovesJesus]

This is great thanks.

[@geospatialschool]

You're welcome!

[@coolsamuel4354]

Massive massive massive