See your dataset clearly, clean it properly
A desktop dataset manager for deep learning. Visual browsing and statistics, annotation quality checks, YOLO / COCO / VOC conversion, and batch label editing — built to work with Easy Trainer Station.
What it does
Each capability maps to a concrete everyday problem.
Visual browsing and stats
Browse dataset distribution by class, size, and annotation count to spot imbalanced classes fast.
Annotation quality checks
Find out-of-bounds boxes, empty labels, duplicates, and tiny targets, then review each flagged sample.
Format conversion
Convert between YOLO, COCO, and VOC while preserving the original image layout and splits.
Batch label editing
Rename, merge, or delete classes in bulk, optionally limited to samples matching a rule.
Class remapping
Unify class tables from multiple sources so the same object does not carry different names.
Noise filtering
Filter by resolution, aspect ratio, or annotation density to produce a clean training set.
How it works in practice
Not a feature list — a walk through getting something done.
See what the dataset looks like before changing it
A thumbnail wall plus class statistics reveals which classes are thin and which sizes look wrong.
Open any sample to see the full image with boxes, and page through with the keyboard.
Flag problem annotations and fix them in bulk
Quality checks list specific issues — out-of-bounds boxes, empty labels, duplicates — so you fix by category instead of re-reviewing everything.
Filtered results can be corrected or dropped in bulk, with a preview of what will change.
Convert to the format you train with
Export the same data as YOLO, COCO, or VOC with a fixed split ratio and seed for reproducibility.
The exported layout is picked up directly by Easy Trainer Station, no manual sorting required.
Design trade-offs
Deterministic work stays in code; judgement stays with you.
Originals untouched
All edits happen on a copy; the original dataset stays intact.
Reproducible conversions
Fixed split ratios and seeds mean the same input always yields the same output.
Consistent toolchain
Shares class and directory conventions with Easy Infer Station and Easy Trainer Station.
Native desktop
Processes large files locally — no dataset uploads anywhere.
What you might want to know first
Does it upload my dataset?
No. Everything runs locally and the app works fully offline.
Which annotation formats are supported?
YOLO, COCO, and VOC, convertible in any direction.
How large a dataset can it handle?
Files are processed individually rather than loaded at once, so datasets with hundreds of thousands of images remain workable.
Start using it now
Free to download, runs locally. Your data stays on your machine.
This is a desktop app — open this page on your computer to download.
Other formats
SHA256 checksum
Provided as-is. The app does not collect usage data.