An open-source pylon components and fault detection dataset for training the ML Algorithms

# About Dataset

This is an open dataset for the pylon components inspection community. The dataset contains 555 images (jpg and png) of three different classes of pylon components (Insulator, Pylon, and Covered Insulator) and their respective annotation (bounding boxes) files. This is a three classes problem and images vary in dimensions and captured from different angles from the drone camera. 

<!-- It's a great first dataset for getting started with pylons inspection. -->

## Dataset Information

    Total Number of Images (jpg, png): 555

        - DJI_*jpg

        ...

        - Frame_*jpg

        ...

        - Image_*png

        ...

    Total Number of annotation files: 555

        - DJI_*txt

        ...

        - Frame_*txt

        ...

        - Image_*txt

        ...

    "Classes.txt" contains the class names according to their index in the annotation files. 

    Classes.txt:

        - Insulator

        - Pylon

        - Covered Insulator

    Each class is represent by index in the annotation files fo related image

        - Insulator -->  0

        - Pylon --> 1

        - Covered Insulator --> 2

## Example

For example for DJI_0001.jpg annotated file DJI_0001.txt contains following information:

0   0.552125    0.273778    0.073250    0.190222

0   0.047875    0.710889    0.033750    0.052889

| Class | X pixel  | Y Pixel | Box Width | Box Height |

|:-:|:-:|:-:|:-:|:-:|

| 0 | 0.552125 | 0.273778 | 0.073250 | 0.190222 |

| 0 | 0.047875 | 0.710889 | 0.033750 | 0.052889 |

where 0 is the index of the class in classes.txt file, the remaining 4 values are bounding box information of normalized [x, y, width, height]

## Credit

These images are taken from the Drone by Oscar Bowen Schofield at SDU UAS center and annotated by Naeem Ayoub. The Dataset is now being released as an open-source for research purposes only.