Skip to content

Latest commit

 

History

History
19 lines (14 loc) · 1.23 KB

README.md

File metadata and controls

19 lines (14 loc) · 1.23 KB
Raw

BBoxerwGradCAM

This class forms boundary boxes (rectangle and polygon) using GradCAM outputs for a given image.

The purpose of this class is to develop Rectangle and Polygon coordinates that define an object based on an image classification model.

The 'automatic' creation of these coordinates, which are often included in COCO JSONs used to train object detection models, is valuable because data preparation and labeling can be a time consuming task.

This class takes 5 user inputs:

  • Pretrained Learner (image classification model)
  • GradCAM Heatmap (heatmap of GradCAM object - formed by a pretrained image classification learner)
  • Source Image
  • Image Resizing Scale (also applied to corresponding GradCAM heatmap)
  • BBOX Rectangle Resizing Scale
  • Class is compatible with google colab and other Python 3 enivronments

update

Weights for Grad-CAM are formed in the final convolutional layer, can a seperate model can be introduced on the Grad-CAM heatmap, or possibly leverage the contouring and gradient strategies used here, to increase IoU after Grad-CAM or CHIP (Channel-wise Disentangled Interpretation of Deep Convolutional Neural Networks)? Relevant Paper: https://arxiv.org/pdf/1902.02497.pdf