InceptionFCN

This is the InceptionFCN network a modified network of the InceptionTime was done as a part of Time Series Classification with InceptionFCN research done by Usmankhujaev et al. The original code was taken from InceptionTime: Finding AlexNet for Time Series Classification (https://github.com/hfawaz/InceptionTime)

Getting Started

First of all you need to install the enviorenments from enviorenments.yml file to your conda enviorenment by running following command

conda env create -f enviorenment.yml

and then

conda activate dl4-tsc

For network training you have to download UCR archive from here (https://www.cs.ucr.edu/~eamonn/time_series_data/) and create the folder named archive and put the UCR archive inside and rename the folder to TSC or you can change your folder names or pathes within the code. Go to main.py file and change the directories in line 78.

Also you can go to /utils/ folder and change the constansts

For network training instructions follow the authors of InceptionTime We implemented the a different network within the existing network named ICNv2.py in the /classfiers/

Network overview

The overview of our Inception (a) block and FCN (b) blocks.

Checking the results

Once you train your network it will save all the metrics results in the separate folders for each class and separate files. You can change the name of the folder you want to save. Here we created results_extractor.py that will help you to extract all the results from all folders and saves it in single .csv file so you can easily see your results.