The module implements a linear classifier wrapped around liblinear library.

Description

This module is responsible for learning and classify feature vectors. Input features are passed, the output are the scores of SVM machines. The module provides training methods with real-time performances.

The commands sent as bottles to the module port /linearClassifier/rpc are the following:

(notation: [.] identifies a vocab, <.> specifies a double, "." specifies a string)

SAVE
format: [save] "class_name"
action: starts to dump feature vector in the "class_name" folder.

STOP
format: [stop]
action: stops all the current activities, the module will be waiting for new commands.

TRAIN
format: [train]
action: starts the training process by reading all the database previously acquired via the [save] command.

FORGET
format: [forget] "class"
action: forgets the "class", deleting all the feature vectors in the database. If "class"="all" all the classes are forgotten.

LIST
format: [objList]
action: retrieves the current classes saved in the database.

CHANGE NAME
format: [changeName] "old_name" "new_name"
action: change the current name of a known object with a new name.

RECOGNIZE
format: [recognize]
action: starts the recognition process.

Libraries

YARP libraries.
liblinear

Ports Created

/linearClassifier/features:i receives feature vector.
/linearClassifier/classification:o streams out the current class winner name.
/linearClassifier/scores:o streams out the scores for the received input in the format "class" "score".
/linearClassifier/rpc receives requests for SIFT dump or dictionary learning.

Parameters

–BufferSize buffer

specify the bufferSize for weighting the output scores.

–CSVM value

specify the value for the regularization parameter used during the training.

–WeightedSVM [0 | 1]

Useful parameter of unbalanced dataset. If 1 it weights positive examples such as their contribution is equal to the negative ones.

–databaseFolder database

specify the database folder were feature vector will be stored.

Tested OS

Linux, Windows 7

Author: Sean Ryan Fanello