Ask Your Question
0

Can we use pretrained TensorFlow model to detect objects in OpenCV?

asked 2017-09-15 13:24:02 -0500

nrparikh gravatar image

The project I am working on has a requirement to detect a person in the frame at a weird angle. I have tried using cascade classifiers but they don't work very well in my case. As a solution to this, I want to use TensorFlow's object detection API specifically import inception model into OpenCV and then do all the processing in OpenCV itself. I have used GOTURN tracker before which uses dnn module of OpenCV and reads the caffee model. I want to do something very similar, has anyone done anything similar to this or can anyone point me in the right direction?

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
2

answered 2017-09-15 23:20:20 -0500

dkurt gravatar image

Mobilenet SSD from Caffe works well, https://github.com/chuanqi305/MobileN.... There are about 20 classes including person. This is a sample: https://github.com/opencv/opencv/blob....

edit flag offensive delete link more

Comments

Thanks...the pointer was really helpful! I have one question though do you know which model has human head/upper body detector?

nrparikh gravatar imagenrparikh ( 2017-09-20 13:54:11 -0500 )edit

Thank you for the quick reply...I did some digging into it and as far as I understand I can import models into OpenCV which can be used to classify images. Then my question is can I use TF object detection API along with OpenCV?

nrparikh gravatar imagenrparikh ( 2017-09-20 14:05:05 -0500 )edit

@nrparikh, I think face detection might be a good estimation of a human head, https://github.com/opencv/opencv/tree.... We are working now to translate it in TensorFlow graph.

dkurt gravatar imagedkurt ( 2017-09-21 00:11:01 -0500 )edit
Login/Signup to Answer

Question Tools

1 follower

Stats

Asked: 2017-09-15 13:24:02 -0500

Seen: 1,335 times

Last updated: Sep 15 '17