Deep learning has revolutionized the realm of computer vision. Neural Networks are widely used in almost all of the cutting-edge tech such as Tesla’s auto-pilot feature. They perform too well that there are times they lead to ethical issues and conflicts. Well, we won’t be diving into those today. Let’s focus on a sub-category of computer vision called “Detection”.
What does one mean by detecting an object? When we see an object, we can exactly point where it is and determine what it is with ease. For computers though, the task is not so simple. This has been an active area of research for years and continues to be so today. In the past decade, with the advent of (rather the resurgence) of deep learning we were able to achieve good results to an extent that it has been possible to use it real-time scenarios.
Here we will be using German Traffic Sign Detection Benchmark(GTSDB) Dataset.
Overview
There are several Neural Network architectures for detection:-
- R-CNN family of architectures
- Single Shot Detectors
- YOLO — You Only Look Once
We will today be seeing the implementation of YOLOv2(A variant of the original YOLO architecture) without going into much details as to how it works.
YOLO came on the computer vision scene with the seminal 2015 paper by Joseph Redmon et al. “You Only Look Once: Unified, Real-Time Object Detection,” and immediately got a lot of attention by fellow computer vision researchers. Here is a TED talk by University of Washington researcher Redmon in 2017 highlighting the state of the art in computer vision.
#tensorflow #gtsdb #yolov2 #object-detection #tensorflow
