MMLA: Multi-Environment, Multi-Species, Low-Altitude Aerial Footage Dataset

Zebras and giraffes at the Mpala Research Centre in Kenya

Example photo from the MMLA dataset and labels generated from model. The image shows a group of zebras and giraffes at the Mpala Research Centre in Kenya.

The MMLA (Multi-Environment, Multi-Species, Low-Altitude Aerial Footage) dataset is a collection of 155,074 frames of low-altitude aerial footage of various species in different environments. The dataset is designed to help researchers and practitioners develop and evaluate object detection models for wildlife monitoring and conservation.

This project provides a fine-tuned YOLO11m model trained on the MMLA dataset, enabling more accurate detection of wildlife in aerial imagery across diverse environments.

Multi-Environment

MMLA includes footage from three distinct locations: Ol Pejeta Conservancy (Kenya), Mpala Research Center (Kenya), and The Wilds Conservation Center (USA).

Multi-Species

The dataset contains annotations for multiple species including Plains zebra, Grevy's zebra, Reticulated giraffe, Masai giraffe, Persian onager, and African wild dog.

Low-Altitude

All footage is captured from low-altitude aerial platforms, providing a realistic dataset for practical wildlife monitoring applications.

Results

Our fine-tuned YOLO11m model achieves the following performance on the MMLA dataset:

Class	Images	Instances	Box(P)	R	mAP50	mAP50-95
all	7,658	44,619	0.867	0.764	0.801	0.488
Zebra	4,430	28,219	0.768	0.647	0.675	0.273
Giraffe	868	1,357	0.788	0.634	0.678	0.314
Onager	172	1,584	0.939	0.776	0.857	0.505
Dog	3,022	13,459	0.973	0.998	0.995	0.860

Dataset Gallery

Representative images from the MMLA dataset showcasing different environments and species:

Ol Pejeta Session 2

Mpala Session 1

The Wilds Session 3

The Wilds Session 1

The Wilds Session 2

The Wilds Session 4

Figure: Representative images from the MMLA dataset showing various environments and species from different collection sessions.

Research Applications

The MMLA dataset facilitates research in several key areas:

Wildlife Conservation

Enables more effective wildlife monitoring in conservation areas through automated detection systems that can process aerial footage.

Computer Vision Development

Provides a challenging benchmark for object detection algorithms in real-world settings with varied lighting, backgrounds, and animal movements.

Population Studies

Facilitates research on wildlife populations through non-invasive monitoring methods that can count and track animal groups.

Behavioral Analysis

Supports studies of animal behavior by providing aerial perspectives that capture group dynamics and movement patterns.

Fine-Tuned Model

We provide a fine-tuned YOLO11m model that has been trained on the MMLA dataset. The model achieves state-of-the-art performance for wildlife detection in aerial imagery.

Full model weights and details are available on HuggingFace.

Dataset Details

The MMLA dataset comprises 155,074 frames from 11 sessions across 3 locations, featuring multiple species of wildlife captured from low-altitude aerial footage.

Location	Session	Date	Frames	Species
Ol Pejeta Conservancy, Kenya	1	1/31/25	16,726	Plains zebra
Ol Pejeta Conservancy, Kenya	2	2/1/25	12,542	Plains zebra
Subtotal:			29,268
Mpala Research Center, Kenya	1	1/12/23	16,891	Reticulated giraffe
	2	1/17/23	11,165	Plains zebra
	3	1/18/23	17,940	Grevy's zebra
	4	1/20/23	33,960	Grevy's zebra
	5	1/21/23	24,106	Giraffe, Plains and Grevy's zebras
Subtotal:			104,062
The Wilds Conservation Center, USA	1	6/14/24	13,749	African wild dog
	2	7/31/24	3,436	Reticulated and Masai giraffe
	3	4/18/24	4,053	Persian onager
	4	7/31/24	506	Grevy's zebra
Subtotal:			21,744
Total:			155,074

Key Dataset Features

Multi-Environment: Data collected across various conservation areas in Kenya and the United States
Multi-Species: Contains footage of multiple species including:
- Plains zebra
- Grevy's zebra
- Reticulated giraffe
- Masai giraffe
- Persian onager
- African wild dog
Low-Altitude: All footage captured from aerial platforms at low altitude to simulate practical wildlife monitoring scenarios
High-Quality Annotations: All frames include detailed bounding box annotations for wildlife detection

The dataset is split into three main collections that are available on HuggingFace:

MMLA Wilds: Footage from The Wilds Conservation Center in the USA
MMLA OPC: Footage from Ol Pejeta Conservancy in Kenya
MMLA Mpala: Footage from the Mpala Research Center in Kenya

Citation

If you use this dataset or model in your research, please cite our paper:

@article{kline2025mmla, title={MMLA: Multi-Environment, Multi-Species, Low-Altitude Aerial Footage Dataset}, author={Kline, Jenna and Stevens, Samuel and Maalouf, Guy and Saint-Jean, Camille Rondeau and Ngoc, Dat Nguyen and Mirmehdi, Majid and Guerin, David and Burghardt, Tilo and Pastucha, Elzbieta and Costelloe, Blair and Watson, Matthew and Richardson, Thomas and Lundquist, Ulrik Pagh Schultz}, journal={arXiv preprint arXiv:2504.07744}, year={2025} }

Acknowledgements

We want to thank Samuel Mutisya and William Njoroge at Ol Pejeta Conservancy for their support in facilitating the WildDrone 2025 Hackathon. We would like to thank Dan Beetem and The Wilds for their support in collecting the data. This work was supported by WildDroneEU (EU Horizon Europe MSCA grant 101071224), the Imageomics Institute (NSF HDR Award 2118240), and ICICLE (NSF grant OAC-2112606).

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.