🎨 Segmentation Overview β–Ό

7349

Total Masks

7349

Images with Masks

30.0%

Avg Foreground

11.8%

Avg Boundary

🎯 Segmentation Task: Pet segmentation using trimap masks with three classes: Foreground (1) - Pet pixels, Boundary (2) - Uncertain pixels, Background (3) - Background pixels.

πŸ” Pixel Analysis β–Ό

Pixel Class Distribution

Distribution of foreground, boundary, and background pixels

Class Distribution Statistics

Mean and standard deviation of each class percentage

πŸ“Š Trimap Analysis:
  • Foreground: Definite pet pixels (class 1)
  • Boundary: Uncertain/transition pixels (class 2)
  • Background: Definite background pixels (class 3)
  • Boundary thickness: Indicates annotation uncertainty

πŸ“ Shape Analysis β–Ό

Shape Metrics

Convexity, compactness, and eccentricity analysis

Convexity

0.500

Ratio of object area to convex hull area. Higher values indicate more convex shapes.

Compactness

0.720

4π×Area/PerimeterΒ². Higher values indicate more compact, circular shapes.

Eccentricity

0.450

Ellipse eccentricity. Higher values indicate more elongated shapes.

πŸ”¬ Shape Metrics:
  • Convexity: Measures how "convex" the shape is (0-1, higher = more convex)
  • Compactness: Measures how "round" the shape is (0-1, higher = more circular)
  • Eccentricity: Measures how "elongated" the shape is (0-1, higher = more elongated)

πŸ”² Boundary Analysis β–Ό

Boundary Thickness

Distribution of boundary thickness values

Boundary Smoothness

Distribution of boundary smoothness values

Boundary Complexity

Distribution of boundary complexity values

Thickness

3.57 px

Average boundary thickness. Thicker boundaries indicate more annotation uncertainty.

Smoothness

0.534

Boundary smoothness (0-1). Higher values indicate smoother, less jagged boundaries.

Complexity

2.29

Boundary complexity. Higher values indicate more complex, irregular boundaries.

πŸ” Boundary Properties:
  • Thickness: Width of boundary region (distance transform)
  • Smoothness: How smooth the boundary contour is (curvature analysis)
  • Complexity: How complex/irregular the boundary is (fractal dimension)

βœ… Quality Analysis β–Ό

Quality Metrics

Overall Quality Score: Good (0.95)
Tiny Foreground (<1%) 14
Large Foreground (>80%) 40
No Boundary 14
Excessive Boundary (>20%) 300
Irregular Shapes (convexity <0.5) 1517
πŸ” Quality Indicators:
  • Tiny Foreground: Masks with very small foreground regions (<1% of image)
  • Large Foreground: Masks with very large foreground regions (>80% of image)
  • No Boundary: Masks without boundary regions (potential annotation issue)
  • Excessive Boundary: Masks with too much boundary region (>20% of image)
  • Irregular Shapes: Masks with very irregular shapes (low convexity)

πŸ–ΌοΈ Interactive Sample Gallery

Sample images with segmentation polygon overlay

πŸ“Έ Interactive Gallery: Select breeds to explore samples with segmentation overlay (● Red: Foreground, ● Orange: Boundary). Hover over masks to see detailed statistics.

🐱 Cats (12 breeds)

Select breeds to display:

🐢 Dogs (25 breeds)

Select breeds to display: