VLM output
A forward-facing dashcam on a bus travels down an urban commercial street in Taiwan, passing a pile of roughly a dozen assorted household garbage bags left uncollected on the sidewalk beside a shuttered storefront at Da Liang Street. The vehicle continues along the same street past parked scooters and shopfronts, with no further waste piles or collection activity visible in the later frames.
- Illegal Dumping DetectedYes
- Waste Bag Count10-12
- Waste LocationSidewalk
- Overflow LevelModerate
- Bin Or Container PresentNo
- Collection Vehicle PresentNo
- Waste Worker PresentNo
- Sidewalk ObstructionPartial
- Waste TypeMixed Household
Customer challenges
Street footage is available but litter in it is never recognised
Scale and type (bulky, piled, scattered) are not graded, so dispatch is mis-sized
Jurisdiction depends on the setting (road, park, bus stop, private land) and is disputed
Illegal dumping is found after the fact with no imagery, time or location evidence
No hotspot or time pattern to target repeat offenders
Customer benefits
Inspection shifts from citizen reports only to 45 buses inspecting daily; proactive coverage up 188–200x
AI assigns the correct agency by setting, reducing cross-agency disputes
Reporting time cut 90%; litter found 60 minutes earlier than manual reports
Passed official on-site verification with 75% L1 recognition accuracy, meeting the acceptance threshold
Cleaning resources allocated by litter scale, avoiding over- or under-dispatch
Imagery, time and location retained for inspection and penalty evidence
Key features
Street Litter Detection (Bus-Mounted Cameras)
Detects street litter from bus-mounted cameras.
Litter Scale Grading (Bulky, Piled, Scattered)
Grades litter scale as bulky, piled, or scattered.
Jurisdiction Assignment by Setting (Road, Park, Bus Stop, Private Land)
Assigns jurisdiction by setting, such as road, park, bus stop, or private land.
Illegal Dumping Hotspot Monitoring & Evidence Package
Monitors illegal dumping hotspots and compiles evidence packages.
Cleaning Dispatch & Case Closure
Dispatches cleaning crews and tracks cases through to closure.
Before & after
| Before | After — with Linker AI Nexa |
|---|---|
| Before 265 fixed sources, about 2.4 reports per day | After — with Linker AI Nexa Up to 50,000 video sources, proactive coverage 188–200x |
| Before Litter reported by citizens hours later | After — with Linker AI Nexa Found 60 minutes earlier, reporting time -90% |
| Before Site visit before dispatch | After — with Linker AI Nexa Detection to dispatch ≤ 10 minutes |
| Before Recognition accuracy unmeasured | After — with Linker AI Nexa L1 accuracy 75%, passed official on-site verification |
| Before Illegal dumping found after the fact | After — with Linker AI Nexa Violations at monitored hotspots -80% |
| Before Jurisdiction disputed case by case | After — with Linker AI Nexa 100% auto-assigned by setting |