

Create an On-Prem Dataset¶

On-Premises datasets provide security and privacy protection for data.

1. Add New On-Prem Dataset with CSV File Upload¶

Prepare a CSV file with a list of URLs as shown below:

login to your TrainingData.io account.
Select "DataSet" tab.
Select "Add New Dataset".
Enter name for new Dataset
Select "Local-Hosted Images"
Upload a csv file that arranges URLs like this

2. Add New On-Prem Dataset by Providing Location to a Folder containing Image Files¶

A dataset can be created using "Upload Folder" if the folder is:

a) mounted as a disk (local disk or network storage), AND b) root of webserver serving dataset is same as the uploaded folder

Add new dataset
Select "Local"
Enter a name for the dataset
Click "Upload Folder"
Select a folder that is root of webserver
In video below root of webserver is folder name "imageserver"
Enter "File System Path" as base URL: example http://localhost:8000/
Select "Submit"

What is recommended directory structure?¶

It is recommended that only one folder is uploaded in a dataset. TrainingData.io application supports multiple recursive folders under a selected folder.

Web Server's root directory MUST BE parent directory of selected folder.

Sample directory structure:¶

imageserver/
├── radiology
│   └── TCGA-CA-G4H2
│       ├── DICOM
│       │   ├── 000045.dcm
│       │   ├── 000046.dcm
│       │   ├── 000047.dcm
│       │   ├── 000048.dcm
│       │   ├── 000049.dcm
│       │   └── 000050.dcm
│       ├── Drive
│       │   ├── av001.jpg
│       │   ├── av002.jpg
│       │   ├── av003.jpg
│       │   ├── av004.jpg
│       │   └── av005.jpg
│       └── test
│           └── test1
│               ├── av001.jpg
│               ├── av002.jpg
│               └── PNG
│                   ├── 00012.jpg
│                   ├── 00013.jpg
│                   ├── 00014.jpg
│                   ├── 00015.jpg
│                   └── 00016.jpg
7 directories, 19 files