Skip to main content
  • EN

Hello,

I encountered an issue while using the RasterObjectDetectorSamplePreparer to prepare samples for the RasterObjectDetectorModelTrainer. My goal is to detect certain objects based on a digital orthophoto image. I created an annotation.txt file using OpenCV and provided the relative paths to the negative samples directory and the annotation file as inputs to the Sample Preparer transformer.

 

However, when running my workspace, I received the following translation log message:

Translation FAILED.

The translation was terminated due to the following feature:

RasterObjectDetectorSamplePreparer_<Rejected> (TeeFactory): RasterObjectDetectorSamplePreparer_<Rejected>: Termination Message: 'RasterObjectDetectorSamplePreparer output a <Rejected> feature. To continue translation when features are rejected, change 'Workspace Parameters' > Translation > 'Rejected Feature Handling' to 'Continue Translation''

The feature on tag `PYOUTPUT' failed to be output by the Bridge.

RasterObjectDetectorSamplePreparer_PythonCaller (PythonFactory): The PythonFactory failed to process the feature.

 

Although I got a Background Description File containing 110 negative samples, I did not receive the positives file. I have already researched this issue by reading questions and articles related to the topic, which led me to use relative paths, but still I don't get results.

If anyone has any insights or solutions regarding this matter, I would greatly appreciate your assistance.

Thank you.

Hi @pej.k​, is there a chance you may be able to provide us with your annotation file and negative samples? The OpenCV transformers, such as this one, are notiriously picky about all the parameters so if something is slightly off it could cause the translation to fail. We'd like to compare our own projects with your data to see might be causing the issue.

If you're unable to share the data with us that's fine too.

 

Thanks!

Hi @pej.k​, is there a chance you may be able to provide us with your annotation file and negative samples? The OpenCV transformers, such as this one, are notiriously picky about all the parameters so if something is slightly off it could cause the translation to fail. We'd like to compare our own projects with your data to see might be causing the issue.

If you're unable to share the data with us that's fine too.

 

Thanks!

Hi @danminneyatsaf​​, I appreciate your response. I attached a zip file containing the negative samples, along with the digital orthophoto image that I used to create the annotation file and annotations.txt, which contains the absolute path. Objects that I want to detect are roofs.

 

Please note that I have utilized a single digital orthophoto image for this task, with only three annotations. This was done as an initial test to determine if any results would be obtained. Subsequently, I plan to create additional annotations.

 

Thank you!

Hi @danminneyatsaf​​, I appreciate your response. I attached a zip file containing the negative samples, along with the digital orthophoto image that I used to create the annotation file and annotations.txt, which contains the absolute path. Objects that I want to detect are roofs.

 

Please note that I have utilized a single digital orthophoto image for this task, with only three annotations. This was done as an initial test to determine if any results would be obtained. Subsequently, I plan to create additional annotations.

 

Thank you!

Thank you for sharing your data with us ! I've tested this alongside a colleague of mine who's more familiar with the transformer and we're confident that something is going wrong here with the transformer.

I submitted a bug report with our developers to get this investigated and will let you know when I've heard back from them on what the cause may be.

For your reference, the ticket number is FMEENGINE-81612.

 

Apologies for the inconvenience.

@pej.k​ the developer looking into this issue has notified me that we're deprecating the RasterObjectDetector set of transformers for 2024.0 due to a number of issues. Bugginess, security vulnerabilities in the library supporting the transformer, and difficulty maintaining the transformer are some notable ones. We don't have plans to replace the transformer.

One of our technical alliance partner's, Picterra, has a platform which is purpose-built for performing object detection with rasters. We recommend looking into a specific service such as this. If you have any questions about their platform we're happy to try and answer them. Here's a link to the product: https://picterra.ch/

 

You can use Picterra for the object detection, ideally through FME so that the data stays within the same workflows that you had built before.

@pej.k​ the developer looking into this issue has notified me that we're deprecating the RasterObjectDetector set of transformers for 2024.0 due to a number of issues. Bugginess, security vulnerabilities in the library supporting the transformer, and difficulty maintaining the transformer are some notable ones. We don't have plans to replace the transformer.

One of our technical alliance partner's, Picterra, has a platform which is purpose-built for performing object detection with rasters. We recommend looking into a specific service such as this. If you have any questions about their platform we're happy to try and answer them. Here's a link to the product: https://picterra.ch/

 

You can use Picterra for the object detection, ideally through FME so that the data stays within the same workflows that you had built before.

@danminneyatsaf​ thank you for the help. I will explore Picterra as suggested.

Reply


Community Stats

30,423
Posts
115,518
Replies
38,130
Members
Terms & ConditionsCookie settings
A product of
Safe Software respectfully acknowledges that we live, learn and work on the traditional and unceded territories of the Kwantlen, Katzie, and Semiahmoo First Nations.