Field segment with Random forest

[kamal preet]

Hi,

I was trying to provide training dataset for wheat, Thistle (weed) and soil as three separate classes. However, the output has only 2 bands. Is there a way that i can get 3 band values for the 3 classes in the output.

Thanks

[pspawar]

Hi,

Could you please provide the screenshot of your output, attributes of training samples? Are you using FIELDimageR-QGIS? Which algorithm? CART or RandomForest?

--
You received this message because you are subscribed to the Google Groups "FIELDimageR: A R package to analyze orthomosaic images from field trials." group.
To unsubscribe from this group and stop receiving emails from it, send an email to FIELDimageR...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/FIELDimageR/669ba49e-14df-46aa-8b8f-8901e7bf5ea1n%40googlegroups.com.

[kamal preet]

Hi,

I am attaching the screenshots that you requested. I am using FIELDimageR-QGIS, and Random forest algorithm.

Thank you.

[pspawar]

Hello! Thank you for providing the screenshot. It appears that there may be an issue with the training samples for weeds; there might not be enough samples, or the size of the polygon created for weeds might be too small for effective training. 

Consider using point features instead of polygons for training samples. The accuracy of the classification depends on both the quantity and quality of your training samples dataset. Accuracy of classification depends on several factors. 

Classifying crops (wheat) and weeds can indeed be challenging, especially with RGB images. You can explore other techniques like deep learning algorithms or object based segmentation for better results.

On Tue, 9 Apr, 2024, 17:10 kamal preet, <kamalpr...@gmail.com> wrote:

Hi,

I am attaching the screenshots that you requested. I am using FIELDimageR-QGIS, and Random forest algorithm.

Thank you.

On Tuesday, April 9, 2024 at 9:42:40 a.m. UTC-5 pspa...@gmail.com wrote:

Hi,

Could you please provide the screenshot of your output, attributes of training samples? Are you using FIELDimageR-QGIS? Which algorithm? CART or RandomForest?

On Sat, 6 Apr, 2024, 10:49 kamal preet, <kamalpr...@gmail.com> wrote:

Hi,

I was trying to provide training dataset for wheat, Thistle (weed) and soil as three separate classes. However, the output has only 2 bands. Is there a way that i can get 3 band values for the 3 classes in the output.

Thanks

--
You received this message because you are subscribed to the Google Groups "FIELDimageR: A R package to analyze orthomosaic images from field trials." group.
To unsubscribe from this group and stop receiving emails from it, send an email to FIELDimageR...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/FIELDimageR/669ba49e-14df-46aa-8b8f-8901e7bf5ea1n%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "FIELDimageR: A R package to analyze orthomosaic images from field trials." group.
To unsubscribe from this group and stop receiving emails from it, send an email to FIELDimageR...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/FIELDimageR/0dfb276c-7131-4d75-acce-233337841e9fn%40googlegroups.com.

[kamal preet]

Okay, sounds good. Thanks for your input!