Skip to content

Data Preprocessing Tab

Use Data Preprocessing when your project still starts from raw videos or messy frame folders.

At a glance

Best for Typical output Usually next
Extracting frames, cropping videos, flattening folders Clean media folders and frame sets Setup & Annotation

What this tab does

You can use this tab to:

  • extract frames from one video or a folder of videos
  • sample a lightweight subset for pilot labeling
  • crop and clean a batch of videos with a shared ROI
  • flatten nested frame folders into one annotation-ready image directory

Main sections

1. Frame Extraction

Use this when you want image frames for annotation or review.

Control What it does
Source video or folder Select a single video or a folder of videos
Output root folder Writes extracted frames into per-video subfolders
Mode stride, random, or interactive
Stride / Sample size Controls how many frames are kept

2. Batch Video Crop & Clean

Use this when all videos need the same crop and cleanup.

Control What it does
Video folder Folder containing the source videos
Output subfolder Destination for processed videos
ffmpeg binary Optional override if ffmpeg is not on PATH
Use CUDA Enables GPU encoding when available
Force new ROI Redraw the crop region instead of reusing the saved one

IntegraPose stores the selected ROI in crop_roi.json so repeated runs stay consistent.

3. Frame Transfer

Use this when frames are spread across nested folders and you want one flat image directory for labeling.

Files are renamed with folder prefixes and numeric suffixes when needed so collisions do not overwrite images.

Raw videos
  -> Data Preprocessing
  -> Setup & Annotation

Practical tips

  • Use this tab before Setup & Annotation so later paths point to clean, stable data.
  • If you plan to use Assisted Pose Curation, extract or organize frames here first.
  • Keep video and folder names descriptive so extracted frames remain interpretable later.