Exploratory Data Analysis (EDA)

12. Exploratory Data Analysis (EDA)#

  • helps you understand dataset

  • detect potential issues

  • and refine preprocessing steps

12.1. Some Common EDS Tasks#

  • View image properties (dimensions, color channels, aspect ratios)

  • Check annotation format (bounding boxes, segmentation masks, class labels)

  • Understand class distribution (balance of object categories)

  • Check for missing or corrupt data (empty annotations, unreadable images)

12.2. Tip: Try to build QA/QC in the pipeline#

Quality assurance and Quality control (QA/QC)

  • Review samples batches of annotated or predicted data

  • Use automated tools to spot common errors

  • Have another person double-check the annotations

Contents