EleFind — YOLOv11 Elephant Detection Model
A YOLOv11 model fine-tuned for detecting elephants in high-resolution aerial and drone imagery, designed to work with SAHI (Slicing Aided Hyper Inference).
Try it in the live demo: EleFind on HuggingFace Spaces
Training Configuration
| Parameter | Value |
|---|---|
| Base model | YOLOv11 (pretrained) |
| Task | Object detection (single class: elephant) |
| Image size | 1024 x 1024 |
| Epochs | 100 (early stopping, patience=20) |
| Batch size | 16 |
| Learning rate | 0.01 |
| AMP | Enabled |
| Augmentation | Mosaic, random augment, erasing (0.4), flip LR (0.5) |
Performance (Test Set — 50 images)
| Metric | Value |
|---|---|
| Precision | 53.2 % |
| Recall | 49.1 % |
| F1-Score | 51.0 % |
| mAP@0.5 | 84.3 % |
| True Positives | 185 |
| False Positives | 163 |
| False Negatives | 192 |
SAHI Configuration
| Parameter | Value |
|---|---|
| Slice size | 1024 x 1024 |
| Overlap ratio | 0.30 |
| Confidence threshold | 0.30 |
| IoU threshold (NMS) | 0.40 |
Training Results
Training curves — loss and metric progression over 100 epochs:
Normalized confusion matrix and Precision-Recall curve (mAP@0.5 = 0.843):
Sample validation predictions:
Usage
With SAHI (recommended)
from sahi import AutoDetectionModel
from sahi.predict import get_sliced_prediction
from huggingface_hub import hf_hub_download
model_path = hf_hub_download(
repo_id="iamhelitha/EleFind-yolo11-elephant",
filename="best.pt",
repo_type="model",
)
model = AutoDetectionModel.from_pretrained(
model_type="yolov8", # SAHI uses 'yolov8' for YOLOv8/v11 models
model_path=model_path,
confidence_threshold=0.30,
device="cpu",
)
result = get_sliced_prediction(
image="aerial_image.jpg",
detection_model=model,
slice_height=1024,
slice_width=1024,
overlap_height_ratio=0.30,
overlap_width_ratio=0.30,
postprocess_type="NMS",
postprocess_match_threshold=0.40,
)
print(f"Detected {len(result.object_prediction_list)} elephants")
With Ultralytics directly
from ultralytics import YOLO
from huggingface_hub import hf_hub_download
model_path = hf_hub_download(
repo_id="iamhelitha/EleFind-yolo11-elephant",
filename="best.pt",
repo_type="model",
)
model = YOLO(model_path)
results = model.predict("aerial_image.jpg", conf=0.30)
Intended Use
Designed for wildlife conservation research — counting and locating elephants in aerial survey imagery.
Limitations
- May not generalise to terrains or lighting conditions outside the training distribution
- Optimised for overhead/nadir aerial views; side-angle photographs will perform poorly
- May miss small or heavily occluded elephants
- False positives can occur on rocks, shadows, or similarly sized objects
Citation
@software{guruge2025elefind,
title = {EleFind: Aerial Elephant Detection using YOLOv11 and SAHI},
author = {Guruge, Helitha},
year = {2025},
url = {https://github.com/iamhelitha/EleFind-gradio-ui}
}
Acknowledgments
@dataset{naude2019aerial,
title = {The Aerial Elephant Dataset},
author = {Naud\'{e}, Johannes J. and Joubert, Deon},
year = {2019},
publisher = {Zenodo},
doi = {10.5281/zenodo.3234780},
url = {https://zenodo.org/records/3234780}
}
@software{jocher2023ultralytics,
title = {Ultralytics YOLO},
author = {Jocher, Glenn and Qiu, Jing and Chaurasia, Ayush},
year = {2023},
version = {8.0.0},
url = {https://github.com/ultralytics/ultralytics},
license = {AGPL-3.0}
}
@article{akyon2022sahi,
title = {Slicing Aided Hyper Inference and Fine-tuning for Small Object Detection},
author = {Akyon, Fatih Cagatay and Altinuc, Sinan Onur and Temizel, Alptekin},
journal = {2022 IEEE International Conference on Image Processing (ICIP)},
doi = {10.1109/ICIP46576.2022.9897990},
pages = {966--970},
year = {2022}
}
Author
Helitha Guruge — Undergraduate Research Project
License
MIT