| --- |
| 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. |
|
|
