Background removal benchmark

Introduction

This repo contains the python code implementing two machine learning based background removal algorithms:

• KDTree based background estimation
• Radial feature extraction

Included files are:

• PYmodule/: the code implementing the background removal algorithms
• script/: the code generating the data to be processed by the background removal algorithms (numpy array)
• demo.ipynb: demonstration of the data processing workflow
• plot.ipynb: figures for data illustration without need for interactive widget support

Installation

This code needs Python version 3.12.

Required modules can be installed with the following conda environment file.

conda create -n environment.yml

Example run with ipython notebook

1. Open script/1-gen_sim_peak.ipynb and follow the cells to generate simulated file.
2. Use demo.ipynb to perform the example material analysis.

• Generating the npy files takes about 10min
• Background removal with KD Tree methods takes ~ 1h

3. Use plot.ipynb to visualize the results.

class on the Data matrix

1. relevant methods of KDTree background removal:

• apply_mask(): set Data elements that are masked to 0

• build: calls node.create_children, assign splitted Data region to node,

• estimate(): calls node.estimate: set node.data (assigned elements) to mean

• random_select_subtract(data3d): set data3d to subtracted results

• constant_subtract(data3d)

• load

• save

2. relevant data of tree:

• tree.tensor/node.data
• tree.mask

TODO

‣潣灭敲獳搭瑡੡