For most independent shooters, bib tagging is still difficult to automate reliably. OCR-based approaches exist, but in practice the errors and cleanup often make them no better than manual tagging. Some commercial services will tag images for you if you upload or send the files, but that adds time and cost.

At larger professional events, matching is often helped by race infrastructure rather than image analysis alone. One example is RFID chips in bibs: when runners cross timing points, the system records their identity and time, and finish-line cameras can associate images with that data. That’s very effective, but it depends on the event’s timing system and isn’t something you can easily recreate on your own.

For a DIY workflow, the realistic options are manual entry or outsourced manual entry. One low-cost method mentioned was using Amazon Mechanical Turk to have workers read bib numbers from batches of images. A simple spreadsheet/database-based workflow can also work if you or participants are willing to tag images.

So: for a free or low-cost setup, manual tagging is the most practical answer; for higher accuracy at scale, you’ll need a paid tagging service or event timing data integration.