Thank you, this was really useful. I find your teaching style really user friendly! I followed the tutorial as I don't have the required extensions in ArcGIS, so doing this in QGIS was great. Do you have any tips on converting the output to vector whilst maintaining the ID classifications? Rather than converting to vector each created polygon just having a random ID which you then have to classify? Also, and this might be an entirely different video, but do you have any info on machine learning? What I mean is this is great to classify pixels, but obviously a reservoir and a river might end up as the same output due to both being same pixel value/blue. But is there a way to teach it to say this is a reservoir, based on pixel values and the shape, as opposed to this being a river, which has same or similar pixel value, but different shape? I haven't used QGIS in a while, so dusting off the cobwebs!
Thanks! So you're looking for object identification with rivers vs lakes. That's really an entirely different (more difficult) problem because the spectral signature alone won't give you that information, you need spatial context and shape and those things are scale dependent. You should be able to convert the raster to vector and maintain the classified identifiers.
@@geospatialschool Yes exactly, I feel like this is more of a deep learning/object detection/machine learning analysis rather than image classification. I will dig into it some more and see what is possible. Thanks
no - its the scp thats fucked really- in uni we have about 4 groups of people - doing the same thing - just different times - its 50/50 if the scp works or not
This is gold. Thank you.
Can we do it with an old satellite image too? to find old wetlands or marshlands that used to exist there in the past.
Thank you, this was really useful. I find your teaching style really user friendly! I followed the tutorial as I don't have the required extensions in ArcGIS, so doing this in QGIS was great. Do you have any tips on converting the output to vector whilst maintaining the ID classifications? Rather than converting to vector each created polygon just having a random ID which you then have to classify? Also, and this might be an entirely different video, but do you have any info on machine learning? What I mean is this is great to classify pixels, but obviously a reservoir and a river might end up as the same output due to both being same pixel value/blue. But is there a way to teach it to say this is a reservoir, based on pixel values and the shape, as opposed to this being a river, which has same or similar pixel value, but different shape? I haven't used QGIS in a while, so dusting off the cobwebs!
Thanks! So you're looking for object identification with rivers vs lakes. That's really an entirely different (more difficult) problem because the spectral signature alone won't give you that information, you need spatial context and shape and those things are scale dependent. You should be able to convert the raster to vector and maintain the classified identifiers.
@@geospatialschool Yes exactly, I feel like this is more of a deep learning/object detection/machine learning analysis rather than image classification. I will dig into it some more and see what is possible. Thanks
Sir spc plugging ki kore download korbo 3.10 e
Sir amar create training Input layer korar poreo training input layer as6ena
thank you sir.
Welcome!
multiprocess: error classification . i swear to god, scp is the most scuffed plugin. maybe not, maybe its qgis itself.
no - its the scp thats fucked really- in uni we have about 4 groups of people - doing the same thing - just different times - its 50/50 if the scp works or not