# Getting the classifier from a data set

## Pre-requisites

Install sklearn 0.17.1:

```bash
sudo pip install sklearn==0.17.1
```

Install Cython:

```bash
sudo pip install Cython
```

Clone the Python-pcl repository from the b-it-bots github page:

```bash
git clone git@github.com:b-it-bots/python-pcl.git
```

Go inside the python-pcl folder and run the following:

```bash
sudo python setup.py install
```

## Save the classifier

Go to the location of the 'train\_classifier.py' file inside your catkin workspace and run that file with the given parameters:

```bash
cd *your catkin workspace*/src/mas_perception/mcr_object_recognition_mean_circle/ros/tools/
python train_classifier.py --dataset *add path to the folder with data to be trained here*
```

## Implement the classifier

Run the following to be able to use the classifier:

```bash
roslaunch mcr_object_recognition_mean_circle object_recognition.launch input_pointcloud_topic:=/camera/depth_registered/points target_frame:=base_link classifier:=classifier
```

To visualize your result on rviz, add the PointCloud2, Marker and MarkerArray feature in the rviz menu.

In the Marker tab, select the "/mcr\_perception/scene\_segmentation/bounding\_boxes" topic.

In the MarkerArray tab, select the "/mcr\_perception/scene\_segmentation/labels" topic.

In the PointCloud2 tab, select "/mcr\_perception/scene\_segmentation/tabletop\_clusters" topic.

Publish the message 'e\_start':

```bash
rostopic pub /mcr_perception/scene_segmentation/event_in std_msgs/String "data: 'e_start'"
```

Publish the message 'e\_add\_cloud\_start':

```bash
rostopic pub /mcr_perception/scene_segmentation/event_in std_msgs/String "data: 'e_add_cloud_start'"
```

Publish the message 'e\_segment':

```bash
rostopic pub /mcr_perception/scene_segmentation/event_in std_msgs/String "data: 'e_segment'"
```

This last one will should enable the visualization of the classifier on rviz.