ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Training/using the BiGG detector

asked 2011-06-15 11:50:01 -0500

NStiurca gravatar image

I am trying to evaluate the REIN framework. As a starting point, I am trying to replicate some of the results presented in REIN - A Fast, Robust, Scalable REcognition INfrastructure . First, I am trying to get the BiGG detector trained and running, but I am having problems. I followed the tutorial on training and using BiGG. I trained the detector on 4 objects using the "manually selected 3D bounding box" method. This created 4 files in the database/models/BiGG/ subdirectory. The files are a few hundred KB each (I don't know if that's expected).

The trouble begins when I try running the BiGG detector. It seems to load all 4 models ok, then notifies that it starts doing detections. Next, the bigg detection nodelet continuously outputs the line Detections: 9 regardless of what's in front of the camera, doesn't publish anything on /bigg/detections, and starts consuming more and more RAM at a rate of ~30-40MB/sec until all the system RAM is depleted.

I think the REIN framework looks very promising, so I would really appreciate help getting some of the detectors such as BiGG running so that I can try REIN out.

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2011-06-17 12:05:29 -0500

Vincent Rabaud gravatar image

REIN is not being developed anymore. We are working on a non-ROS specific version of it. Previous object recognition pipelines (like BIGG) will be ported back to it once stable. A release is expected for July/August 2011.

edit flag offensive delete link more

Question Tools


Asked: 2011-06-15 11:50:01 -0500

Seen: 541 times

Last updated: Jun 17 '11