Warning 003387 Input raster cell size error

[jesusangel]

I downloaded sar images and processed them using arcgis, created a data set and trained the U-net neural network, but I have an error when using the "classify pixels using deep learning" tool as I get the following error:

Warning 003387 Input raster cell size 0.000105 is not within the cell size range [0.000109, 0.000110] the Deep Learning model has been trained for. Results may not be optimal. Consider setting another cell size in the environment, or using a different raster input, or using another Deep Learning model.

The sar image in which I want to do the classification has the same characteristics as the one with which I created the data set, I do not understand the error, how can I solve it? since if I run the classification it does not detect anything, it helps

[MohsenLotfi]

Hello dear friend I also encountered such a problem. Were you able to solve this problem? If possible, please guide me as well.

[TexasCityPlanner]

Bumping this topic. Need guidance on the same issue - raster cell size is not within cell size range the Deep Learning Model has been trained for

😞 

[KitWilson]

Me too. What cell size is the model trained for and how should the raster be formatted to make the tool work?

[JoshBerna]

That just means the model extrapolated an expected cell size within that range from training. The cell size of the raster on which you're conducting inference (0.000105) is plenty close enough. 

If it were a more substantial difference (if the model expected 0.005 for instance) you would have to use a different raster for inference that's within that expected cell size, resample and lose visual fidelity, or train a new model with input training raster/s aligned to the cell size you seek. The models all have a certain cell size that they were trained on and they expect that resolution in rasters you pass through them. 

I got this issue on a model I just trained but the expected cell size is 0.003m. This makes no sense to me as the training raster is 0.01m. Came here looking for an explanation for that but I can offer my two sense on your issue for anybody else looking.