topic Re: Deep Learning prepare_data in ArcGIS API for Python Questions

Deep Learning prepare_data

LaurynasGedminas2 — Sun, 12 Dec 2021 09:32:31 GMT

Hello,

I am following the example here for pixel classification:

Pixel-based Classification Workflow with | ArcGIS for Developers

In my case I am exporting data and labels from ArcPro, when i run:

data = prepare_data(data_path, batch_size=16)

I get error:

TypeError: int() argument must be a string, a bytes-like object or a number, not 'NoneType'''

What is the parameter settings needed in ArcPro to export data for this example?

API version: 1.7.0

ArcGIS Pro 2.4

Thank you,

Re: Deep Learning prepare_data

LaurynasGedminas2 — Mon, 25 Nov 2019 19:17:00 GMT

If i export data as 'labelled tiles' i get error:

lib\site-packages\arcgis\learn\_data.py in prepare_data(path, class_mapping, chip_size, val_split_pct, batch_size, transforms, collate_fn, seed, dataset_type, resize_to, **kwargs)    357     358     if dataset_type is None and not has_esri_files:--> 359         raise Exception("Could not infer dataset type.")    360     361     if dataset_type != "Imagenet" and has_esri_files:Exception: Could not infer dataset type.

Re: Deep Learning prepare_data

Anonymous User — Wed, 27 Nov 2019 13:17:13 GMT

Try specifying the dataset_type. So for Classification tiles, add dataset_type='Classified_Tiles' to call the prepare_data function.

Re: Deep Learning prepare_data

LaurynasGedminas2 — Wed, 27 Nov 2019 19:23:50 GMT

when i try to export 'Classified Tiles' thru export tool it gives error: ERROR 002091 Parameters are inconsistent.

Re: Deep Learning prepare_data

Anonymous User — Fri, 29 Nov 2019 19:10:43 GMT

Laurynas Gedminas Try to follow the methodology documented in this link to export training data for classified tiles here Extracting Building Footprints From Drone Data | ArcGIS for Developers .

Re: Deep Learning prepare_data

Anonymous User — Mon, 02 Dec 2019 08:01:46 GMT

Can you post a screenshot of parameters you have supplied to the tool.

Re: Deep Learning prepare_data

Anonymous User — Mon, 02 Dec 2019 08:03:04 GMT

Laurynas Gedminas‌ can you post the full error

TypeError: int() argument must be a string, a bytes-like object or a number, not 'NoneType'''

Re: Deep Learning prepare_data

LaurynasGedminas2 — Mon, 02 Dec 2019 17:38:48 GMT

this gives the ERROR 002091 Parameters are inconsistent.

it's 3 band raster, TIFF or PNG still the same, sites were produced by the Training Sample Manager:

Re: Deep Learning prepare_data

Anonymous User — Tue, 03 Dec 2019 05:42:20 GMT

Can you check if there is any missing values in the ClassValue column .

Re: Deep Learning prepare_data

stephenescarzaga — Wed, 22 Jan 2020 21:20:24 GMT

I just went through this problem. Documentation says that exporting "classified_tiles" requires a classified raster (thematic) instead of feature class. So I had to do polygon to raster, then 'set raster properties'. Once I did that It let me export. Not sure if this is the correct way.

EDIT: looking at the 'extract building foot prints' here, my workflow with ultimately not work. I'm trying to prepare the data now for training and I'm getting errors.

Re: Deep Learning prepare_data

Anonymous User — Thu, 13 Feb 2020 04:43:23 GMT

You can follow the steps documented here https://developers.arcgis.com/python/sample-notebooks/extracting-building-footprints-from-drone-data/