Branch: master
Find file History
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
data
libs
trained_models/OPF_debug
README.txt
README_all.txt
RegressionTubularityAC.pro
RegressionTubularityAC_plugin.cpp
RegressionTubularityAC_plugin.h

README.txt

# This plugin implements the regression tubularity measure described in [1] for 3-Dimensional image stacks.
# Features are computed using predefined convolutional filters that are provided with the plugin.
# Multiscale prediction is not yet supported.

# Training a model from terminal:
./vaa3d -x RegressionTubularityAC -f train
     -i <Train_Img1.tif> <Train_Img2.tif> ... <Train_ImgN.tif>
     -o <Output_Results_Dir>
     -p <GroundTruth_Img1.swc> <GroundTruth_Img2.swc> ... <GroundTruth_imgN.swc>
   	<Path_to_sep_filters_file_im.txt> <Path_to_weights_file_im.txt> <Path_to_sep_filters_file_ac.txt> <Path_to_seigths_file_ac.txt>
        <N_Autocontext_Iters> <N_Train_Samples> <N_boosting_Iters> <Tree_Depth> <Shrinkage_factor> <m_try>

Input: Supported formats: Tiff and vaa3d raw UINT8 3d volumes
Output: The trained models will be saved in <Output_Results_Dir> under the names Regressor_ac_0.cfg, Regressor_ac_1.cfg, ..., Regressor_ac_<N_Autocontext_Iters>.cfg
Parameters: <GroundTruth_Img1.swc> <GroundTruth_Img2.swc> ... <GroundTruth_imgN.swc> swc files contining ground truth. Must be same number of input images.
											 N.B. It Assumes swc files are in image coordinate !
<Path_to_sep_filters_file_im.txt> <Path_to_weights_file_im.txt> <Path_to_sep_filters_file_ac.txt> <Path_to_seigths_file_ac.txt> text files containing separable filter banks and weights to extract features from the image and the tubularity score respectively. You can use the filters provided under the ../data/filter_banks directory

            <N_Autocontext_Iters> number of Auto-context iterations (set to 0 in order not to use Auto-context). Default 1.
            <N_Train_Samples> number of training samples. Default 100000.
            <N_boosting_Iters> number of boosting iterations. Default 200.
            <Tree_Depth> max depth of trees used as weak learners. Default 2.
            <Shrinkage_factor> shrinkage factor. Defualt 0.1.
            <m_try> number of features randomly sampled to lean a weak learner. Default 0 ( = use all features).


# Tesing a model from terminal:
./vaa3d -x RegressionTubularityAC -f test
     -i <Test_Img1.tif> <Test_Img2.tif> ... <Test_ImgN.tif>
     -o  <Tubularity_img1.raw> <Tubularity_Img2.raw> ... <Tubularity_ImgN.v3draw>
     -p <Path_to_sep_filters_file_im.txt> <Path_to_weights_file_im.txt> <Path_to_sep_filters_file_ac.txt> <Path_to_seigths_file_ac.txt>
	<Path_to_Regressor_ac_0.cfg> <Path_to_Regressor_ac_1.cfg> ... <Path_to_Regressor_ac_M.cfg>

Input: Supported formats: Tiff and vaa3d raw UINT8 3d volumes
Output: format: vaa3d raw FLOAT32 3d volumes
Parameters: <Path_to_sep_filters_file_im.txt> <Path_to_weights_file_im.txt> <Path_to_sep_filters_file_ac.txt> <Path_to_seigths_file_ac.txt> text files containing separable filter banks and weights to extract features from the image and the tubularity score respectively. You can use the filters provided under the ../data/filter_banks directory

<Path_to_Regressor_ac_0.cfg> <Path_to_Regressor_ac_1.cfg> ... <Path_to_Regressor_ac_M.cfg> correspond to M+1 trained models, one for each Auto-context iteration.
    When M>0, results for the different Auto-context iterations will be saved in the /temp_results directory



[1] A. Sironi, E. Türetken, V. Lepetit and P. Fua. Multiscale Centerline Detection, submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence.