02-preprocessing-template.md

SCRUM-XX: Preprocessing Pipeline for

Dataset

• Name:
• Source:
• Size: <e.g., 300GB, 440k videos>
• License:

Storage Location

• Path:
• Access Method: <download/access command>

Data Split

• Train/Val/Test Ratio: <e.g., 70/15/15>
• Counts:
  • Train:
  • Val:
  • Test:
• Method: <how split was created, including seed/stratification and grouping rules>

Preprocessing Steps

1. <step 1>
2. <step 2>
3. <step 3>

Code Changes

• Updated <path/to/file>
• Added <path/to/file>
• Modified <path/to/file>

How to Run

# Step 1: Download/locate dataset
<command>

# Step 2: Split dataset
<command>

# Step 3: Run preprocessing
<command>

Verification

• Ran preprocessing end-to-end
• Verified split counts match expectations
• Verified processed data looks correct (sample check)
• No dataset files committed
• Documentation updated

Notes

<gotchas, assumptions, and important details for teammates>