Computer vision for retail, built rack-first.
GoTrack is a multi-stage computer-vision platform that turns every clothing rack and shelf into a sensor — detecting pickups, identifying the exact variant, triggering signage, and producing consideration analytics no POS system can.
Retail computer vision is the application of machine perception to physical stores — detecting shopper interactions with fixtures, identifying products without barcodes, and turning camera streams into the consideration data that online retail has always had and offline retail has never had.
Every spoke below is a slice of the same pipeline: vision-AI detection, multi-object tracking, fashion-aware recognition, catalog matching, multi-signal ranking, rack state machine, signage trigger. One platform, every retail surface.
See the platformPick a problem. Pick a vertical. Same stack.
Pickup detection
The core capability — see what shoppers actually take off the rack, identify the exact variant.
Shrinkage reduction
Self-checkout label-switching, abandoned-pickup patterns, evidence trails for LP teams.
Planogram optimization
Rack-position heatmaps from real pickup data. A/B test layouts like e-commerce A/B tests pages.
Consideration analytics
Hold time, pickup-to-purchase ratio, cross-product journeys — consideration data that doesn't exist anywhere else.
Digital signage trigger
Camera detects a pickup, the matching content plays on the right LCD within milliseconds.
Visual inventory
Empty-fixture detection, variant-level stock signals, restock alerts from cameras instead of RFID.
Fashion retail
Fashion-aware recognition outperforms generic models on apparel. Built for 10,500-variant catalogs.
Luxury & boutique
Discretion-first observation. Hand-feel time as the highest-intent purchase signal.
Department stores
One stack, every department. Per-floor analytics. Cross-department journey mapping.
Mall operators
Multi-tenant analytics. Common-area traffic plus per-store engagement, in one dashboard.
Three numbers we ship in every store.
Sub-300ms hot path
Camera frame to LCD swap, end-to-end, at p95. Measured in production, not a synthetic benchmark.
Fashion-aware accuracy
Fashion-domain recognition, multi-frame confirmation, and a high-confidence match threshold. Generic models guess; our recognition AI knows.
Drop-in on existing IP cameras
ONVIF-compatible cameras, no rip-and-replace. Four hours to deploy per store, including the LCD pairing.
What a 4-week pilot looks like.
A mid-tier Turkish apparel group rolled GoTrack into a single flagship store. Four cameras, four LCDs, one Sunday-evening install. By week two the LP team had identified a label-switch pattern that POS data had missed for three quarters. By week four merchandising had moved two slow SKUs to higher-pickup rack positions — both saw double-digit lift the next inventory cycle.
We had heard about pickup data for years. The first dashboard view, on a Monday morning, showed us a slow-moving product getting picked up fifty times for every sale. Three weeks later we'd renegotiated the price. Nothing else has ever shown us that.
Frequently asked
Do we need to replace our existing cameras?
No. GoTrack runs on any ONVIF-compatible IP camera that's already pointed at a rack or shelf. 1080p at 10 FPS is enough; higher resolution is welcome but not required.
How is this different from generic computer vision?
Two things. First, our fashion-aware recognition AI — trained specifically on fashion data — outperforms generic recognition models, which catastrophically fail on garments. Second, a multi-stage pipeline with multi-frame confirmation, evidence accumulation, and a rack state machine eliminates the false triggers that kill generic CV in retail.
How long does a pilot take?
Four hours to install hardware in a single store. Two to four weeks to gather enough pickup data to talk about it. Twelve weeks to integrate signage triggers, planogram changes, and LP workflows.
What about shopper privacy?
GoTrack never stores faces, never identifies individuals, and operates on local network where possible. We track garments and tracks (anonymous IDs that expire on exit), not people. KVKK and GDPR compliant by design.
Does it work on grocery or electronics?
The pipeline is generic; the recognition is fashion-specific. For non-apparel we either retrain our recognition on category-specific data or pair GoTrack with a category-appropriate model. Fashion is where we are strongest today.
Can we run it offline?
Yes. The pickup detection and signage trigger run entirely on local network. Cloud sync (for analytics aggregation and multi-store dashboards) is opt-in and resilient to outages.
Stop guessing what your shoppers consider.
Online retail has clickstream. Physical retail has the parking lot, the door, and the receipt — and seven hours of darkness in between. GoTrack fills the dark. Camera ingestion to LCD swap in under 300 milliseconds, consideration analytics that no POS can produce, and a 4-hour deploy. Book a demo and we'll bring the rack to you.