4,500 runners received race photos matched by bib number and face recognition, all delivered on the same day with an average processing time of 4.2 seconds per photo.
The Mubadala Abu Dhabi Marathon is one of the Gulf region's largest running events, attracting 4,500 participants across marathon, half-marathon, and 10K distances. With photographers stationed at 12 points along the course — including start line, key kilometre markers, water stations, and the finish — the event generates roughly 38,000 photos in a single morning.
Historically, race photos were uploaded to a third-party gallery where runners had to manually scroll through thousands of images to find their own shots. The experience was frustrating: runners spent 20 to 30 minutes searching, many gave up before finding their photos, and the race organisers received hundreds of support emails asking for help locating specific images.
The organising committee wanted every runner to receive only their own race photos, delivered automatically to their phone, ideally before they left the venue. The system needed to handle the unique challenges of marathon photography: runners in motion, similar-looking kit, partially obscured faces, and bib numbers that are sometimes folded, covered by timing belts, or splashed with water.
Eventiere deployed a dual-matching approach specifically designed for running events. Two AI systems ran in parallel on every uploaded photo: bib number detection using optical character recognition tuned for race bibs, and facial recognition for high-accuracy face matching even with sunglasses, caps, and sweat-streaked faces.
Runners registered before race day through the event's existing registration flow, which included a selfie capture step. On race morning, photographers uploaded images in batches every 10 to 15 minutes using mobile hotspots along the course. Each batch was processed within seconds — faces detected, bib numbers read, and photos matched to runner profiles.
When both bib and face matched the same runner, confidence was highest. When only one signal was available — a distant shot showing a clear bib but no face, or a close-up finish line photo with no visible bib — the system still made accurate matches using the single available identifier. Results were delivered via WhatsApp, with runners receiving a notification each time new photos of them were found.
"Our runners were sharing their race photos on Instagram before they'd even finished their cool-down. The speed was genuinely impressive. We've never had this kind of instant social media buzz from race day."
All 38,000 photos were processed and matched with an average delivery time of 4.2 seconds per photo from upload to appearing in a runner's gallery. By the time the last finishers crossed the line, early starters had already received multiple batches of their race photos spanning the full course.
The dual bib-plus-face approach achieved a 97% match rate across all photos where at least one identifier was visible. Support emails related to missing or mismatched photos dropped from over 300 in previous years to fewer than 15. Social media posts tagging the marathon increased by 4x on race day, with runners sharing their professional race photos directly from their Eventiere galleries.
The organising committee calculated that eliminating the manual photo-sorting process — which had previously required a team of six people working for three days post-race — saved over 140 person-hours. Runner satisfaction surveys showed a 92% approval rating for the photo delivery experience, making it the highest-rated aspect of the event experience alongside the course itself.