Double Gross / Double Tare
Same tractor-trolley weighed twice in a single transaction window — or tared with a different vehicle than the one that was grossed.
A computer-vision augmentation for sugar mill weighbridges. Detects malpractice in real time, captures 60-second video evidence around every flagged transaction, and routes alerts into the channels your supervisors already check.
Medium-scale co-operative and private sugar mills in India. Currently working with a federation operating 23+ factories — with scale potential to 40–50 mills as adoption expands.
Manual weighbridge operations at factory yards lose significant value to malpractice and inefficiency that’s difficult to detect after the fact. Most are pattern-based and only visible in real time — not in a post-season audit.
Same tractor-trolley weighed twice in a single transaction window — or tared with a different vehicle than the one that was grossed.
Front wheel crossing the platform edge, off-center loading, vehicle straddling two bridges. All distort the recorded weight.
Driver still seated on the tractor, or a person standing on the platform during weighing — either adds tens of kilograms per transaction.
Weight blocks, mud, fallen cane, ropes — quietly added to the trolley between gross and tare to inflate the recorded cane weight.
A heavier vehicle category logged as a lighter one in the weighment software — changes the trolley tare baseline and inflates net cane.
Disputes happen weeks after the fact. By then the footage is overwritten and the trail is cold. Farmer grievances stay unresolved.
We don’t require new hardware to start. The system runs on existing yard infrastructure with optional upgrades planned in later phases.
Currently 2 cameras per bridge (front + side). Optional overhead camera in later phases for top-down trolley inspection. Day + night with IR.
Real-time triggers from Gross / Tare transactions in the weighment system the factory already runs. AI receives the event and the transaction metadata as it happens.
Day and night footage from factory yards across the federation. Used both for model training and for evidence playback on flagged transactions.
Platform ROI, vehicle types, human-presence cases, foreign objects, positioning categories — annotated from real yard footage. Plus vehicle and transaction metadata from the current systems.
Three phases. The MVP ships in 4 weeks at one or two weighbridges — enough to demonstrate end-to-end value before scale.
4 weeks — 1–2 weighbridges. Foundation work + core detections: platform ROI, vehicle movement, vehicle type, position alerts, human detection. End-to-end pipeline live.
Next 8–12 weeks — Add foreign-material detection and advanced detections. Scale to 5–8 factories. Tune false-positive rate on real-world data.
6–9 months total — Complete rollout across 23+ factories with monitoring dashboards, alert routing, and federation-level reporting. Go-live target: before next crushing season.
Clear targets at MVP and at full rollout. Both technical metrics and operational ones — engineered systems should commit to both.
We’ll bring an engineer on-site, audit your existing cameras and weighment integration, and have an MVP running on one weighbridge inside a month.