sandbox338/wildlife-detector-detectron2

🐾 Wildlife Detector (Detectron2)

A fine-tuned Faster R-CNN object detection model trained with Detectron2 to identify 8 species of wild animals from real-world images. This model is suitable for conservation applications, automated image tagging, and wildlife monitoring.

📦 Model Details

• Model Type: Faster R-CNN (ResNet-101 + FPN)
• Framework: Detectron2 (Facebook AI)
• Task: Object Detection (bounding box)
• Classes: Antelope, Lion, Elephant, Zebra, Gorilla, Wolf, Leopard, Giraffe

📈 Evaluation

Metric	Score
mAP (0.5:0.95)	70.0
AP50	97.8
AP75	84.3
AP (small)	10.2
AP (medium)	54.0
AP (large)	70.8

Per-Class Performance

Class	AP (%)
Antelope	67.4
Lion	68.8
Elephant	78.3
Zebra	66.7
Gorilla	70.8
Wolf	72.1
Leopard	72.1
Giraffe	64.0

🧪 How to Use

To run inference:

from detectron2.engine import DefaultPredictor
from detectron2.config import get_cfg
from detectron2 import model_zoo

cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file("COCO-Detection/faster_rcnn_R_101_FPN_3x.yaml"))
cfg.MODEL.WEIGHTS = "path/to/model_final.pth"
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 8

predictor = DefaultPredictor(cfg)
outputs = predictor(your_image)