# YOLOExplorer

A Python tool for exploring and iterating on computer vision datasets with SQL filters, vector similarity search and a GUI dashboard, built for Ultralytics YOLO and related models.

YOLOExplorer lets a user analyze, filter and clean computer vision datasets with a simple Python API: it supports SQL filters, vector similarity search, and native interoperability with Pandas. It helps find and remove problem images such as duplicates and out-of-domain data, and can enrich a dataset by pulling in examples from another dataset.

A GUI dashboard supports exploring multiple datasets at once, searching across them, and moving images between them to fix weak spots, with quick access to start training on a newly built dataset. Multiple pretrained feature-extraction models, including ResNet, EfficientNet, GoogLeNet and MobileNet variants, can be used to build embeddings for a dataset.

It can rapidly generate new dataset versions trainable on Ultralytics YOLO, SAM, FAST-SAM and RT-DETR, storing embeddings in a LanceDB table for querying. It installs with pip.

Website: https://github.com/lancedb/yoloexplorer