Skip to content

matatusko/Real-Time-Object-Recognition-in-Keras

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 

Repository files navigation

Real time object recognition in Keras

Very simple model built with Keras (tensorflow backend) and computer vision using pretrained weights from ResNet50. Mostly built for fun to play around with computer vision.

The project is hugely inspired by and adopted from Chun's Machine Learning Page, where the whole process is explained very neatly, so I would hugely recommend the lecture.
To be more specific, the two following blog posts:
Real Time Object Recognition (Part 1)
Real Time Object Recognition (Part 2)

Nevertheless, some of the code provided by Chun did not work for me and I had to do some adjustments. So if you struggle with his code, maybe my implementation will help you out. It works well with Keras v2.0.0 and tensorflow v1.4 at the backend of the whole operation.

As you can see with the examples below, it works all right. Most of the everyday objects get recognized correctly, however sometimes you can see silly mistakes like my automatic pencil being classified as a revolver with nearly 50% prediction accuracy, which is ResNet50's pretrained weights fault. Nevertheless, I suppose that makes it funnier to play around with :)

Example

alt text

About

Real time object recognition in Keras based on ImageRes

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages