Object detection using Intel OpenVINO, Tensorflow on the IDD dataset
This test run is the executed by recording from a dashcam of low quality.
Please ignore the date of the video frame as shown. My dashcam has a battery problem so when it is, the timer is also off.
The training is done using Tensorflow Object Recognition pipeline and Mobilenet SSD was used as the initial checkpoint.
It was done on a CPU and thats why the detection is very bad.
After training, Intel Distribution for OpenVINO toolkit was used to optimize it and ultimately for inference as shown in the video.
Subsequent runs will be executed and it will be a part of the VAMS project.
IDD dataset link: https://idd.insaan.iiit.ac.in
VAMS: https://software.intel.com/en-us/arti...
Code and article will be released soon on medium after more 3-4 test runs
Music: / bells