Pickup an object

Introduction

This package is based on the object-recognition package (see the References part). It allows to detect custom objects from a pre-trained network, using the bag-of-words methods, detect its pose on a table, thanks to OpenCV, and then grasp it.

How-to...

Launch the node

Open a terminal, and execute the following command: roslaunch object_recognition demo.launch

Train on new objects

According to the project Git repo (see References for further information):

1. Browse to src/object-recognition/src/bag_of_words
2. Capture images from the Baxter's arm camera, thanks to the baxter_capture_images.py script (check the output folder location!)
3. Then use the sift_train_classifier.py script to train the classifier (check the output .pkl file location!)
4. Check which weights file is used for the object recognition, in the baxter_object_prediction.py file (line 185)

References

• object-recognition project from apollack11 : https://github.com/apollack11/object-recognition