Automatically detect, through camera and Artificial Intelligence (AI), whether a shopping cart is full or empty at the moment of passing through the POS.
The system generates real-time alerts in Darwin (via the CARRINHO_CHEIO rule), integrating the EASiCash video layer with operational and anti-fraud data.
- The camera captures a lateral (90°) internal view of the shopping cart.
- The video stream is sent directly to a GPU-enabled server (no intermediary NVR).
- The AI processes frames in real time and classifies the cart as full or empty.
- A Darwin alert is generated, with logs available for auditing and BI.
| Model |
Type |
Resolution |
Notes |
| DS-2CD2121G0-I(W)(S) |
Dome |
2 MP |
Project standard. Wi-Fi and audio versions |
| DS-2CD2021G1-I |
Bullet |
2 MP |
Fixed-lens alternative |
| DS-2CD2143G2-I(S) |
Dome |
4 MP |
Higher image quality |
| DS-2CD2347G2-LU |
Turret |
4 MP |
High performance in low-light environments |
| DS-2CD2746G2-IZS |
Varifocal Dome |
4 MP |
Adjustable for wide areas |
⚠️ Important:
- The camera must be connected directly to the NVR/DVR or to the GPU server, on the same switch.
- No mirroring, multiple paths, or intermediary routers are allowed.
- Cameras routed through NVRs before reaching the server may fail detection.
- Height: Between 1.70 m and 1.90 m
- Angle: 90° lateral to the cart path
- Environment: Stable lighting, no reflections
The analysis is performed locally, on a server equipped with a dedicated RTX GPU, within the same network as the EASiBox.
- Motherboard: Gigabyte B760M DS3H Rev. 1.0
- CPU: Intel Core i7-12700K
- RAM: 16 GB DDR4 3200 MHz
- SSD: 1 TB NVMe – Kingston
- GPU: NVIDIA RTX 4060 or higher (minimum 8 GB GDDR6 with Ray Tracing support)
- Cooler: DeepCool AK400
- Power Supply: Corsair CX Series 750W
- Case: DeepCool CH360
- Motherboard: Same as Option 1
- CPU: Intel Core i5-13600KF
- Other components: Identical to Option 1
- Motherboard: ASUS TUF Gaming B550-PLUS
- CPU: AMD Ryzen 7 5800X
- RAM / SSD / GPU / PSU / Cooler: Same as Intel configuration
- Motherboard: Same as AMD Option 1
- CPU: AMD Ryzen 7 5700X3D
- Other components: Same as the previous option
- DeepCool CH360
- Phanteks XT Pro
- Cooler Master MasterBox NR400
🚨 For stable operation:
- The camera and the server must be on the same physical switch
- No VLANs, routers, or multiple paths between the video source and destination
- Network latency must be minimal. Do not use Wi-Fi, low-performance switches, or intermediary links
🔧 This step is mandatory and critical for proper analytics operation.
- Cameras must be fully operational in EASiCash
- The engineering team remotely accesses the server
- Data capture and model training are performed per store
- The process takes approximately 15 days after installation
- After validation, the AI goes into production with alerts in Darwin
🛑 The model is trained specifically per store and is not reused across locations.
- Alert rule:
CARRINHO_CHEIO
- Visualization and filtering available in the client dashboard via
https://darwin.easi.live
- Data can also be correlated through BI or D+1 auditing
▶️ Watch video
▶️ Watch video
▶️ Watch video