This repository was developed to assist biologists at analyzing large image collections from chimpanzees (e.g., created with camera traps or by Zoo visitors). However, algorithms are not specifically tuned to chimpanzees and are also applicable to other species.

The provided pipeline is intended to tackle a variety of questions, including

• How many chimpanzees? (Overall statistics)
• Where? (Detection)
• Who? (Identification + Novelty Detection)
• Which Age? (Regression)
• Which Age Group? (Classification)
• Which Gender? (Classification)

Below is a possible output of our entire pipeline.

• the source code was developed in Matlab 8.2
• to use all currently supported features, the following libraries are required
  • Darknet (object detection)
  • Caffe (feature extraction, classification, regression)
  • LibLinear (classification)
  • gpml (regression)
• For a full list, see Installation guidelines below.

• clone this repository (git clone https://github.com/cvjena/analyzing-chimpanzees)
• run initWorkspaceChimpanzees.m (libraries not found on your system are listed and possible code mirrors are suggested)
• Download the following repos, install them, and add their paths in initWorkspaceChimpanzees.m (thereby we make their locations known to Matlab)
  • CaffeTools (https://github.com/cvjena/caffe_tools)
  • LibLinearWrapper (https://github.com/cvjena/liblinearwrapper)
  • gpml (http://www.gaussianprocess.org/gpml/code/matlab/doc/)
  • Darknet (https://github.com/cvjena/darknet)
  • Chimpanzee Face Dataset (https://github.com/cvjena/chimpanzee_face)
• Note that CaffeTools and LibLinearWrapper come with their own initWorkspace scripts, which also need to be adapted (require the correct pointers to Caffe and LibLinear at your system)
• Check that all 3rd-party libraries are installed andspecify their location in initWorkspaceChimpanzees.m
• Run initWorkspaceChimpanzees.m again.

After you set up paths to 3rdparty libraries (see Installation), you can start to explore the repository! Before running the demos, make sure to download some pre-trained models from download-from-google-drive and to copy them to repo/demos/

• Analyzing cropped face images (identification, age, age group, gender)
  • run demo_czoo_face_analysis.m
  • note: requires pre-trained models
• Perform an entire evaluation on CZoo (identification, age, age group, gender)
  • run demo_czoo_evaluation.m
• Visualize GT data on non-cropped images
  • run demo_showing_ground_truth_data.m
• Detect chimpanzees in non-cropped images
  • run demo_only_detection.m
  • note: requires pre-trained models
• Detect chimpanzees in non-cropped images + analyze detected faces
  • run demo_detection_and_face_analysis.m
  • note: requires pre-trained models

This software is licensed under the non-commercial license Creative Commons Attribution 4.0 International License. For usage beyond the scope of this license, please contact Alex Freytag.