🧩 1. How a Convolution Layer Works

• When you apply a convolution layer on an image:
  • Example: 6 filters of size 5×5×3 on an image of size 32×32×3.
  • Output: 28×28×6.
• In code, you stretch the filter then slide it over the image: 5×5×3 = 75 parameters per filter.

➡️ 2. Stride

• Stride controls how far the filter moves across the image.

• It determines how many pixels the filter skips each step.

• Formula: Example: For (7−3)/2 + 1 = 5.

  Output Size = ((Input Size - Filter Size) / Stride) + 1
  
  

🔳 3 & 4. Padding

• Zero Padding adds extra pixels around the image edges to keep output size under control.

• New formula:

  Output Size = ((Input Size - Filter Size + 2 × Padding) / Stride) + 1
  
  

⚠️ 5. Shrinking Effect

• Each added convolution layer shrinks the image.
• Too many layers can mean losing valuable information.

🧮 6. Number of Parameters

• Example: For one convolution layer:
  • Filter: 5×5×3 = 75
  • +1 (bias) = 76
  • For 10 filters: 76 × 10 = 760 parameters.