Data preparation

Date

2019-04-03

This part introduce how to prepare and load data for the project

Generate Signals and Annotations

This project is the simplified version of classification and detection 2d signals. So the first thing we should do is generate some 1d data to feed the algorithm. the data package is created for this purpose

Load Data from files

After data generation, we will get a bunch of signal files and a json file that contains the annotation information for each of these signals. This annotation file follows the convention of COCO dataset 1 which is widely used in object detection and segmentation researches.

Pytorch DataLoader

Pytorch is very convenient and efficient in data loading, there’s a package called DataLoader which can help you load data from files with multi-cpu parallel processing. As long as you give it the proper method of loading a single dataframe, it will help you load the whole dataset in a very short period of time.

References

1

An Image Dataset with abandant annotations which is created for advanced image classification, object detection and segmentation researches. See COCO