bsridatta / Rotated-Object-Detection

Tiny ResNet inspired FPN network (<2M params) for Rotated Object Detection using 5-parameter Modulated Rotation Loss

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

rotated_ship_data is not found

xing-shuai opened this issue · comments

there is no rotated_ship_data.py under src folder

Hello @xing-shuai, thanks for your interest!
Yes, unfortunately, this repo does not contain the dataset. I do not have access/rights to the dataset I worked with.
However, to fill in the information gap, the network is made to take an image and perform a 'single rotated object', 'single class' detection.
The focus of the work is on exploring Modulated Rotated Loss and how it compares to normal L1, L2 losses, how it behaves with other activations functions etc. And also to explore small and efficient architectures.

The idea is to scale it to multiple objects and multi-class to support something like this dataset
https://www.kaggle.com/rhammell/ships-in-satellite-imagery/code