Python engine report viewer

Audit image datasets before they enter modeling.

Import `latest-report.json` from the Python engine, or run a browser-only Fast Scan for structure and class balance.

Quality score
0

Run Python Audit Engine first

The browser cannot execute Python. Run this command in your local terminal, then import the generated `latest-report.json` here.

pnpm engine:scan -- --path ./dataset --out ./reports/latest-report.json

No Python report imported yet. Deep evidence cards are intentionally marked as unchecked.

Import latest-report.json

Load the JSON report after the Python engine finishes. This unlocks leakage, duplicate, corrupt image, blur, brightness, and contrast evidence.

Browser Fast Scan

Quickly parse local folders for labels, splits, dataset size, and class-balance risk. It does not check duplicate, leakage, corrupt, or blur evidence.

Risk Breakdown

Leakage

requires train/val/test structure and exact hash comparison.

Not checked
Not checked

Duplicate

SHA-256 duplicate detection runs in the Python audit report.

Not checked in Fast Scan
Not checked in Fast Scan

Imbalance

Choose a folder to parse class distribution.

Not checked
Not checked

Resolution

Import a Python report for full image dimensions and low-resolution rate.

Not checked in Fast Scan
Not checked in Fast Scan

Blur

Laplacian blur score needs the Python audit engine.

Not checked in Fast Scan
Not checked in Fast Scan

Brightness / Contrast

Brightness and contrast statistics need the Python audit engine.

Not checked in Fast Scan
Not checked in Fast Scan

Leakage requires train/val/test folders. A train-only scan has no validation or test split to compare against.

Class distribution

Import a Python audit report or choose a local folder for Browser Fast Scan. The dashboard will render class counts, risk evidence, and deterministic recommendations here.

Evidence cards

LeakageNot checked in Fast Scan

Needs train/val/test hash comparison.

Duplicate groupsNot checked in Fast Scan

Exact SHA-256 duplicate detection runs in Python audit.

Blurred imagesNot checked in Fast Scan

Laplacian blur score is part of the Python report.

Corrupt imagesNot checked in Fast Scan

Decoded locally by the Python image engine.

Recommendations

low

No scan evidence loaded yet.

Evidence: The dashboard has not received latest-report.json or Browser Fast Scan results.

Action: Run the Python engine or choose a local image folder to start the audit.