📦 Detection Overview ▼

3671

Total Bounding Boxes

3671

Images with Annotations

1.0

Avg Boxes per Image

1

Object Class (Pet Head)

🎯 Detection Task: Pet head detection using bounding box annotations in Pascal VOC format. Each image contains one pet head with precise bounding box coordinates.

📏 Bounding Box Properties ▼

Size Statistics

Mean and median width/height of bounding boxes

Aspect Ratio Distribution

Distribution of wide, square, and tall bounding boxes

Area Distribution

Distribution of normalized bounding box areas

Size Categories

COCO-style size categorization (small/medium/large)

📊 Key Insights:
  • Size Categories: Based on diagonal length (small: <32px, medium: 32-96px, large: >96px)
  • Aspect Ratios: Wide (>1.5), Square (0.8-1.2), Tall (<0.67)
  • Normalized Areas: Bounding box area relative to image area

🗺️ Spatial Distribution ▼

Position Heatmap

Density map of bounding box center positions

Center Bias Analysis

Distribution between center and edge regions

3x3 Grid Distribution

Distribution of bounding boxes across image regions

🎯 Spatial Analysis:
  • Center Bias: Tendency for objects to appear in image center
  • Position Heatmap: Visual density of bounding box centers
  • Grid Analysis: Distribution across 3x3 image regions

✅ Quality Analysis ▼

Quality Metrics

Overall Quality Score: Excellent (1.00)
Invalid Boxes 0
Tiny Boxes (<100px²) 0
Large Boxes (>90% image) 5
Extreme Aspect Ratios 0
🔍 Quality Indicators:
  • Invalid Boxes: Negative dimensions or zero area
  • Tiny Boxes: Too small for reliable detection
  • Large Boxes: Cover most of the image (potential annotation error)
  • Extreme Ratios: Very wide or very tall boxes (ratio >10 or <0.1)

🖼️ Interactive Sample Gallery

Sample images with bounding box annotations overlaid

📸 Interactive Gallery: Select breeds to explore samples with bounding box overlay (Green: Pet head detection).

🐱 Cats (12 breeds)

Select breeds to display:

🐶 Dogs (25 breeds)

Select breeds to display: