Package Summary

Tags No category tags.
Version 1.0.0
License LGPL
Build type CATKIN
Use RECOMMENDED

Repository Summary

Checkout URI https://github.com/ipa320/cob_environment_perception.git
VCS Type git
VCS Version indigo_dev
Last Updated 2018-10-04
Dev Status UNMAINTAINED
Released UNRELEASED

Package Description

PCL feature evaluation.

Additional Links

Maintainers

  • Georg Arbeiter

Authors

  • Georg Arbeiter
The following explains the evaluation process and use of the different program parts and scripts

___Labeling________________________________________________________________________________________

From raw point clouds to ground-truth
___________________________________________________________________________________________________

  - pcd_to_ppm:
    extract the depth and rgb data of pcd files

  - extract_organized_curvature:
    extract min and max curvature of pcd files to ppm color image to make labeling easier

  - Import rgb, depth and curvature images in GIMP to label manually

  - ppm_to_pcd:
    map back the labeled ppm file onto the point cloud

___Best Configuration Parameter____________________________________________________________________

Determing the best configuration fo RSD and PC for each scene
___________________________________________________________________________________________________

  - generate_test_data:
    generate feature values (RSD,PC,FPFH) for a list of normal and feature radius parameters

  - evaluate_test_data:
    evaluate the previously generated list of feature values with several config parameters of
    RSD and PC. the results are saved for all combinations in a .csv file

  - use OpenOffice Calc to find the best combinations

___FPFH Training___________________________________________________________________________________

Setting up SVM for FPFH classification
___________________________________________________________________________________________________

  - generatePrimitives.sh + fpfh_primitives:
    generate FPFH feature values for convex and concave synthetic shapes in different sizes for
    all classes

  - extractFeatureValues.sh + extract_feature_values:
    extract FPFH feature values for all classes from the previously generated test data using the
    manually labeled scenes

  - reduceFeatureValues.sh + reduce_fpfh_training_data:
    perform k-means to select the most discriminating histograms (about 1,000 from over 800,000) 
    from all available data to reduce training time of SVM

  - trainSVMs.sh + fpfh_svm_trainer:
    run OpenCV svm trainer in autotrainig mode using the reduced set of FPFH feature values

  - evaluateFPFH.sh + feature_evaluation_fpfh:
    evaluate several configuration combinations of fpfh and create .csv file
    use OpenOffice Calc to find the best one

___Testing Results_________________________________________________________________________________

Get a look on the results for the best configuration
___________________________________________________________________________________________________

  - feature_evaluation:
    take raw and labeled pcd, set up the desired configurations and visualize the classification
    results for all features

___________________________________________________________________________________________________


CHANGELOG
No CHANGELOG found.

Wiki Tutorials

See ROS Wiki Tutorials for more details.

Source Tutorials

Not currently indexed.

Package Dependencies

System Dependencies

No direct system dependencies.

Dependant Packages

No known dependants.

Launch files

No launch files found

Messages

No message files found.

Services

No service files found

Plugins

No plugins found.

Recent questions tagged cob_3d_evaluation_features at answers.ros.org

Package Summary

Tags No category tags.
Version 1.0.0
License LGPL
Build type CATKIN
Use RECOMMENDED

Repository Summary

Checkout URI https://github.com/ipa320/cob_environment_perception.git
VCS Type git
VCS Version hydro_dev
Last Updated 2014-06-18
Dev Status UNMAINTAINED
Released UNRELEASED

Package Description

PCL feature evaluation.

Additional Links

Maintainers

  • Georg Arbeiter

Authors

  • Georg Arbeiter
The following explains the evaluation process and use of the different program parts and scripts

___Labeling________________________________________________________________________________________

From raw point clouds to ground-truth
___________________________________________________________________________________________________

  - pcd_to_ppm:
    extract the depth and rgb data of pcd files

  - extract_organized_curvature:
    extract min and max curvature of pcd files to ppm color image to make labeling easier

  - Import rgb, depth and curvature images in GIMP to label manually

  - ppm_to_pcd:
    map back the labeled ppm file onto the point cloud

___Best Configuration Parameter____________________________________________________________________

Determing the best configuration fo RSD and PC for each scene
___________________________________________________________________________________________________

  - generate_test_data:
    generate feature values (RSD,PC,FPFH) for a list of normal and feature radius parameters

  - evaluate_test_data:
    evaluate the previously generated list of feature values with several config parameters of
    RSD and PC. the results are saved for all combinations in a .csv file

  - use OpenOffice Calc to find the best combinations

___FPFH Training___________________________________________________________________________________

Setting up SVM for FPFH classification
___________________________________________________________________________________________________

  - generatePrimitives.sh + fpfh_primitives:
    generate FPFH feature values for convex and concave synthetic shapes in different sizes for
    all classes

  - extractFeatureValues.sh + extract_feature_values:
    extract FPFH feature values for all classes from the previously generated test data using the
    manually labeled scenes

  - reduceFeatureValues.sh + reduce_fpfh_training_data:
    perform k-means to select the most discriminating histograms (about 1,000 from over 800,000) 
    from all available data to reduce training time of SVM

  - trainSVMs.sh + fpfh_svm_trainer:
    run OpenCV svm trainer in autotrainig mode using the reduced set of FPFH feature values

  - evaluateFPFH.sh + feature_evaluation_fpfh:
    evaluate several configuration combinations of fpfh and create .csv file
    use OpenOffice Calc to find the best one

___Testing Results_________________________________________________________________________________

Get a look on the results for the best configuration
___________________________________________________________________________________________________

  - feature_evaluation:
    take raw and labeled pcd, set up the desired configurations and visualize the classification
    results for all features

___________________________________________________________________________________________________


CHANGELOG
No CHANGELOG found.

Wiki Tutorials

See ROS Wiki Tutorials for more details.

Source Tutorials

Not currently indexed.

Package Dependencies

System Dependencies

No direct system dependencies.

Launch files

No launch files found

Messages

No message files found.

Services

No service files found

Plugins

No plugins found.

Recent questions tagged cob_3d_evaluation_features at answers.ros.org