ccf_person_identification

This package is an implementation of person identification based on the combination of Convolutional Channel Features and Online boosting. It takes advantage of deep feature representation while adapting the person classifier to a specific target person.

Example

Test images can be found in "ccf_person_identification/data/test".

rosrun ccf_person_identification ccf_person_identification_test

Training images

Positive images

Negative images

Test images

confidence: 0.545271, -0.545632, -0.540605

Extracted and selected features

Related packages

• ccf_person_identification
• monocular_people_tracking
• monocular_person_following

Papers

• Kenji Koide, Jun Miura, and Emanuele Menegatti, Monocular Person Tracking and Identification with Online Deep Feature Selection for Person Following Robots, Robotics and Autonomous Systems [link].

• Kenji Koide and Jun Miura, Convolutional Channel Features-based Person Identification for Person Following Robots, 15th International Conference IAS-15, Baden-Baden, Germany, 2018 [link].