YOLO Dataset Merge Tool
Drag & drop two YOLO datasets. Class conflicts are handled automatically, then normalized and downloaded in one click.
The ZIP must contain images/ and labels/ folders, plus a data.yaml with the class names list · No sign-up · Data cleaned up right away
Free resource hosted for the community. On the first merge the server may need a moment to wake up (~30s) — please wait.
Why merge YOLO datasets in the browser?
Training a robust model usually means combining detection data from several sources — but each dataset often uses its own class names and ID numbers. YOLO Merge handles that work for you: it pairs the classes between your two datasets, detects same-name conflicts, renumbers everything into one consistent order, and outputs a new data.yaml so the merged dataset drops straight into your training script.
How it works
1 · Upload two datasets
Drag & drop two YOLO dataset ZIPs. Each one is scanned and its classes, image count and labels are shown.
2 · Review class mapping
Duplicate class names are detected. Adjust the merge order, rename classes, or click a name to split them — the final ID for every class is generated from this order.
3 · Download & train
Get a single merged ZIP with images/, labels/ and a freshly generated data.yaml, ready for YOLO training.
Frequently asked questions
›What must the ZIP file contain?
Each dataset ZIP needs an images/ and a labels/ folder, plus a data.yaml that lists the class names. Both folders are matched by file name — the same basename in each, with a .jpg/.png and a .txt label respectively.
›How are class ID conflicts handled?
When the two datasets use different numbers for the same class, YOLO Merge deduplicates by name and renumbers everything in a single, conflict-free order. You can preview and adjust the final order, or rename any class, before merging.
›Is it free? Why is there a wait on my first use?
Yes — it is completely free. The backend runs on a free hosting tier that sleeps when idle, so the very first request may take about 30 seconds to cold-start. A progress bar appears while it wakes up; subsequent actions are fast.
›Is my data private?
Your ZIPs are uploaded only to process the merge. No account or sign-up is required, and the temporary session (and the uploaded files) are deleted from the server when you leave the page.
›Which annotation format does it support?
YOLO TXT labels (.txt with class center_x center_y width height, normalized coordinates). Merged datasets drop-in for Ultralytics YOLO training with a regenerated data.yaml.