Skip to content

taimur1871/image_heatmaps

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Using Activations of Pretrained Network to Create Image Heat Maps

This workbook explores the idea of generting heat maps from a pretrained model to create heat maps that allow for rough detection of object in an image. An example of dog detection using this is given below.

Original

heatmap

Enironment

  • python 3.6
  • tensorflow 2.3
  • matplotlib 3.3.3
  • PIL 8.0.1
  • numpy 1.19
  • cv2 4.4.0.46

Extension of Application

The idea can be taken much further and used to generate bounding boxes. This was done by making a threshold map from heatmaps and then finding coordinates of the object of interest. Please refer to notebook 3 for some examples. Threshold_heat_map

This resulted in the following bounding box dog_bbox

The examples below are some of the tests carried out on other animals.

Ferret Ferret_pred

Raccoon Raccoon_pred

Rabbit Rabbit_pred

Mouse Mouse_pred

About

Using pre-trained models to create crude object detection models

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published