• Home
  • Projects
  • Playground
  • Scribbles
  • CV
Colorlib 2018 @ All rights reserved
<

Image by Towards Data Science

Artificial Intelligence

>Realtime Object Recognition

  • Nov, 2019

While taking a famous (as all others start with) coursera's Machine Learning course, I decided to explore some of the open source API with more practical applications. The first attempt was to run real-time object recogntion in PC environment.
I referred to lots of detailed steps from towardsdatascience.com

DEVELOPMENT

Environment : MacOS, Tensorflow CPU
Resource : Tensorflow Object Detection API / OpenCV / Python 3.6
Pre-trained model : detection model ZOO trained on COCO Dataset

Google's Object Recognition API provides most of the code to train/evaluate/deploy customized recognition models farely easily. The most of the time will be taken in the process of taking various picture in different environment, occluded with different objects etc. and labeling those pictures manually.
The detailed process can be found in this Youtube video

MY RESULT

Environment : MacOS, Tensorflow CPU
Resource : Tensorflow Object Detection API / OpenCV / Python 3.6
Pre-trained model : detection model ZOO trained on COCO Dataset

The first video is tested with SSD MobileNet with Tensorflow CPU installed in MacOS, and the second one with Yolo v2 with Tensorflow GPU (CUDA 9.0/CUDNN 7.0) with Surfacebook 2.