How to Resolve Single Shot Detector(SSD) issues

1861
8
03-07-2020 10:48 PM
ManojG
by
New Contributor II

Hi,

I am new to Deep Learning and arcgis.learn module. I am following the case study "Palm Tree Detection" . While Running the code the below Error is occurring. How can I Resolve this Issue?

arcgis.learn object detection‌deep_learning arcgis-api-python palm tree-detection‌

ssd‌ single-shot-detector‌ #fastai

0 Kudos
8 Replies
DavidPike
MVP Frequent Contributor

From a purely python perspective it's not understanding your class name 'models'.

Try pytorch.models.resnet34 or whatever module you imported.

ManojG
by
New Contributor II

Thank you David.

But I Just imported SingleShotDetector function from arcgis.learn. Which is having the modules from pytorch, fastai and scikit-image

0 Kudos
DavidPike
MVP Frequent Contributor

If it's just being passed as a parameter, try = 'RESNET34'

0 Kudos
ManojG
by
New Contributor II

I tried this as you suggested. But the Errors raised at the time of importing packages. so I Undo the changes

0 Kudos
DanPatterson_Retired
MVP Emeritus

Did you set up your environment properly?

Use Deep Learning to Assess Palm Tree Health | Learn ArcGIS 

There are issues if you don't have the correct versions set up

For example

Deep Learning to Assess Palm Tree Health: .emd file error 

Deep Learning to Assess Palm Tree - prepare_data problem 

plus many more.

You might want to look at their github site for various examples and check to see if anything there helps

arcgis-python-api/samples/04_gis_analysts_data_scientists at master · Esri/arcgis-python-api · GitHu... 

0 Kudos
ManojG
by
New Contributor II

I am following the link(https://learn.arcgis.com/en/projects/use-deep-learning-to-assess-palm-tree-health/) from first onwards. 

I removed the palm-detection environment.

again created the environment and same packages installed. 

This time the error happening in preparedata. 

The ArcGIS pro is 2.5 version

Arcgis Python api is 1.7.0 version

the H/W is intel  i7-8th generation with NVIDIA GeForce GT 730(Desktop)

The process is in Deadlock.

One thing I really didn't get.

I already ran this case study in my laptop (i7-8th generation and switchable graphics[AMD Radeon]) with 20 epochs.

It ran smoothly. Now I would like to increase the number of epochs to 50. This will take much time to complete the training. So, this time I go with Desktop. Nothing Changed in the process of execution. But the errors are occurring about the Package installations.

How can I resolve this issue?

Are there any Hardware Setting changes needed?

0 Kudos
DanPatterson_Retired
MVP Emeritus

Contact the people that wrote the lesson (If they are listed) or tech support, there seems to be a lot of fiddly stuff with those deep learning and related modules

by Anonymous User
Not applicable

Manoj have you installed the required dependencies for arcgis.learn. Can you paste the output of command "conda list" here.

0 Kudos