All Projects
Computer Vision

YOLOv8 Animals Detection

80-class animal detection with YOLOv8n. mAP@0.5=0.668, mAP@0.5:0.95=0.560. Best: Tiger (0.967), Sparrow (0.953). Challenging: Squid (0.009). ONNX (12.3 MB). 29,071 images across 80 species.

0.6681
mAP@0.5 (overall)
0.967
Best class (Tiger)
0.5596
mAP@0.5:0.95
3.0ms / image
Inference speed
Dataset

22,566 images, 80 species, CVAT annotated

Approach

CVAT→YOLO conversion → YOLOv8n fine-tuning → ONNX export for edge deployment

Tech Stack
PythonYOLOv8 (Ultralytics)ONNXCVATAdamW
Keywords
YOLOv8Object Detection80-classWildlifeONNXEdge Deployment
Visualizations5 Charts
Deep Dive

80-class animal detection pipeline using YOLOv8n.

Dataset

  • 22,566 annotated images: 18,053 train / 4,513 val + 6,505 test
  • 80 animal species: Bear → Zebra (alphabetically)
  • CVAT XML → YOLO normalized format (multi-threaded, 4 CPU workers)
  • Class imbalance: 2 images (rare species) to 100+ (common species)

Training YOLOv8n (3.15M params, 73 layers) — 50 epochs, batch=16, imgsz=640, AdamW (lr=1e-3), patience=10

Overall Performance

MetricValue
mAP@0.50.6681
mAP@0.5:0.950.5596
Precision0.6706
Recall0.6478
Inference speed3.0ms / image

Per-Class Highlights

Best SpeciesmAP@0.5Worst SpeciesmAP@0.5
Tiger0.967Squid0.009
Sparrow0.953Seahorse0.000
Rhinoceros0.940Hamster0.230
Hedgehog0.942

Performance variance is almost entirely driven by training sample count. The 3.0ms inference speed makes this viable for real-time wildlife monitoring applications.

Deployment: ONNX export (12.3 MB) for edge devices.