gorilla-watch/Gorilla-Zoo-Berlin

Gorilla-Berlin-Zoo Dataset

Part of the GorillaWatch project — a cross-domain evaluation benchmark for gorilla re-identification in a controlled zoo environment.

Dataset Description

The dataset contains 154 videos of 5 individual Western Lowland Gorillas (Gorilla gorilla gorilla) recorded across 3 cameras over 3 months at Berlin Zoo. It provides 188,692 annotated face bounding boxes across 275 tracklets, making it suitable for re-identification, face/body detection, and cross-domain generalization research.

Unlike in-the-wild datasets such as Gorilla-SPAC-Wild, this dataset features consistent camera placement, controlled lighting, and zoo-specific domain properties (glass, artificial structures, distinct backgrounds).

Individuals

Name	Notes
Bibi
Tilla
Djambala
Sango
M'Penzi

Cameras

zoo1 · zoo2 · zoo3

Configurations

Config	Description	Samples
face_with_body	Face crop paired with full body crop	188,679
body	Body crop only	401,785
full_image_bbox_body	Full video frame + body bounding box	401,785
full_image_bbox_face_with_body	Full video frame + face and body bounding boxes	184,626

Metadata Schema

All configs share these fields:

Field	Type	Description
image	Image	Main image (face crop, body crop, or full frame depending on config)
class	string	Gorilla identity
date	string	Recording date (YYYY-MM-DD)
time	string	Recording time (HH:MM:SS)
video	string	Source video filename
frame_number	int	Frame index within the video
camera	string	Camera identifier

Config-specific fields:

Config	Extra fields
face_with_body	body_image — paired body crop
full_image_bbox_body	bbox — body bounding box [x, y, w, h]
full_image_bbox_face_with_body	bbox — face bounding box [x, y, w, h], body_bbox — body bounding box [x, y, w, h]

Raw Videos

The original 154 MP4 recordings are bundled as videos.tar.gz alongside the parquet shards. Extract with:

tar -xzf videos.tar.gz

Usage

from datasets import load_dataset

# Face crop paired with body crop (default)
ds = load_dataset("gorilla-watch/Gorilla-Zoo-Berlin", "face_with_body")

# Full frame with face + body bounding boxes
ds = load_dataset("gorilla-watch/Gorilla-Zoo-Berlin", "full_image_bbox_face_with_body")

Performance Benchmarks

Results from our paper on the test split:

Method	Strategy	Top-1 Accuracy
Ensemble	Confidence Averaging	84.75%
Ensemble	Embedding Averaging	80.61%
InternVideo2	—	65.09%
TimeStormer	ViT	64.59%
AIM	ViT	53.56%

Ensemble methods significantly outperform end-to-end video architectures on this benchmark.

Citation

@inproceedings{GorillaWatch2026,
  title     = {GorillaWatch: An Automated System for In-the-Wild Gorilla Re-Identification and Population Monitoring},
  booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  author    = {Maximilian Schall and Felix Leonard Knöfel and Noah Elias König and Jan Jonas Kubeler and
               Maximilian von Klinski and Joan Wilhelm Linnemann and Xiaoshi Liu and
               Iven Jelle Schlegelmilch and Ole Woyciniuk and Alexandra Schild and
               Dante Wasmuht and Magdalena Bermejo Espinet and German Illera Basas and Gerard de Melo},
  year      = {2026},
  archivePrefix = {arXiv},
  eprint    = {2512.07776}
}

License

CC BY 4.0

Acknowledgments

We are grateful to Zoo Berlin for their expert assistance and facility access, enabling the development of tools to support gorilla conservation.