/LikelihoodRatio_ML

Primary LanguageJupyter Notebook

LikelihoodRatio_ML

Introduction

Results

Future work

Centered bounding boxes

In order to improve the model, we need to have more bounding boxes in locations other than the center of the image. To do this, we could obtain the RA/Dec coordinates of the bounding box and create subimages containing the bounding box, but have it moved off center in the subimages. This would likely help to remove the bias in the model to place boxes primarily in the center of the image, leading to better results.

Box selection

Use IR components to produce training data

Other box definition

$(X center,Y center, width, height)$