---
pretty_name: Birds Object Detection 1600x896
task_categories:
  - object-detection
size_categories:
  - 10K<n<100K
tags:
  - birds
  - yolo
  - ultralytics
  - object-detection
---

# Birds Object Detection 1600x896

A single-class bird object-detection dataset prepared for Ultralytics at a
1600 x 896 camera resolution. The dataset uses a deterministic 90% training
and 10% validation split.

## Contents

- `data/yolo_birds_1600x896.tar.gz`: complete Ultralytics dataset archive.
- `manifest.json`: archive size, SHA-256 checksum, and dataset metadata.

After extraction, the archive contains `data.yaml` and matching image and YOLO
label trees:

```text
data.yaml
images/train/
images/val/
labels/train/
labels/val/
```

## Download

```python
from huggingface_hub import hf_hub_download

archive = hf_hub_download(
    repo_id="PBatch23888/birds-object-detection-1600x896",
    repo_type="dataset",
    filename="data/yolo_birds_1600x896.tar.gz",
)
```

Private repositories require an authenticated Hugging Face account with
access to the dataset.

## Dataset composition

The data builder combines bird annotations from Open Images, COCO 2017,
Pascal VOC 2012, Birdsnap, and NABirds. Open Images group-of and depiction
annotations are excluded. All retained annotations are mapped to the single
class `bird`.

## Licensing

This is a compilation of multiple upstream datasets. Their respective terms
and licenses continue to apply. Review each upstream dataset's license before
redistributing or using this compilation.