MMLA Pose: Eight-Way Orientation Labels for Zebras in Drone Footage
- Provider
- Imageomics Institute (Sun C., WildWing team)
- License
- CC0-1.0 (source)
- Size
- 1,026 files
- Format
- PNG, JPEG (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/imageomics/mmla-pose)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Computer Vision
- Source
- https://imageomics.github.io/wildwing/
About MMLA Pose: Eight-Way Orientation Labels for Zebras in Drone Footage
988 manually labelled 224x224 RGB crops of zebras taken from low-altitude drone footage at the Mpala Research Centre and Ol Pejeta Conservancy in Kenya and The Wilds conservation centre in Ohio, each assigned one of eight pose-orientation classes: front, front-left, left, back-left, back, back-right, right, front-right. The label says which flank faces the camera, which matters for stripe-based individual re-identification and for autonomous drone positioning in the WildWing system (YOLO detector plus a DINOv2 + MLP viewpoint classifier). Classes are imbalanced: 369 right and 295 left side views against 44 to 84 for the oblique and frontal classes; 35 ambiguous crops are kept in a _skip folder. The layout is folder-per-label with filenames encoding source session, DJI video, partition, frame and box index, so each crop traces back to the MMLA and KABR source videos. 1,026 files, about 20 MB; load with torchvision ImageFolder or datasets imagefolder. Dedicated to the public domain under CC0 1.0; cite the dataset and KABR.
How to load MMLA Pose: Eight-Way Orientation Labels for Zebras in Drone Footage in Python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="imageomics/mmla-pose", filename="_reference.png", repo_type="dataset")
print(path)Data structure
File types in sampled files: PNG, JPEG
Files
Example file: _reference.png
Analyze this dataset without downloading it
On Scigantic, MMLA Pose: Eight-Way Orientation Labels for Zebras in Drone Footage opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Computer Vision datasets
- Vipe Dynpose 100 Kpp
- Llama Nemotron VLM Dataset V 1
- IntPhys 2
- TartanAir V2: multi-sensor synthetic dataset for SLAM, navigation and robot perception
- Nemotron Research Lgt
- Vipe Wild Sdg 1 M
- GeckoNum: Numerical Reasoning Benchmark for Text-to-Image Models, Generated Images from Imagen, Muse and DALL-E 3 with Counting Prompts (Google DeepMind)
- Nemotron Image Training V 3