Certificate ID: GSF-SCI-2026-00002
DWaste
DWaste Platform
v1.4.0
Self-certified compliant with ISO/IEC 21031:2024
Software Carbon Intensity (SCI)
Issued: November 3, 2026
Expires: November 3, 2027
About the Software
DWaste is an edge-AI and IoT-enabled platform for sustainable waste intelligence, automated recycling sorting, and real-time contamination tracking serving educational institutions, businesses, public spaces, and Material Recovery Facilities (MRFs). The platform deploys lightweight computer-vision models directly on smartphones and edge devices including full offline functionality to classify waste into seven recycling-critical categories (biological, cardboard, glass, metal, paper, plastic, trash).
SCI Score
Organization
DWaste
Measurement Period
October 22, 2026 – October 22, 2026
Issued
November 3, 2026
Software Boundary
Included Components
| Component | Description | Justification |
|---|---|---|
| Edge Inference Runtime | Quantized YOLOv8n-CBAM model execution layer (3.5 MB footprint) |
Excluded Components
| Component | Description | Rationale for Exclusion |
|---|---|---|
| Cloud Analytics Backend | Excluded because core classification operates cloud-free by design; optional external syncing falls outside the measured operational boundary. | |
| Model Training Phase | Excluded from operational SCI per ISO/IEC 21031:2024 guidelines (one-time development cost of 0.09458 kgCO2, reported separately). | |
| Network Transmission | Excluded by architectural design as all image inference and auditing occur locally on-device. | |
| Actuator Control Signals | Excluded because the control signals that drive the sorting hardware (6-compartment flapper bins and conveyor machines) are emitted as low-power digital GPIO/PWM outputs from an embedded microcontroller. Their operational energy is negligible relative to the vision inference runtime, and they execute on a separate hardware path (microcontroller) outside the measured edge inference boundary. | |
| On-Device Vision Pipeline, Mobile & Audit App Runtimes | Excluded because these components execute on separate hardware paths (mobile CPU/ISP/DSP and embedded microcontroller) that are outside the measured edge inference runtime boundary. This is a deliberate boundary decision: the SCI assessment isolates the vision inference runtime as the sole attributable operational component under DWaste's control, and treats the surrounding capture, display, and audit application layers as host-platform functions. | |
| Local Storage | Excluded because the 3.5 MB ONNX model is memory-mapped and read once per inference; estimated read energy is ~0.005 J (using a conservative NAND benchmark of ~1.5 J/GB), which is under 0.3% of the measured 1.85 J inference energy (5.15 × 10⁻⁷ kWh). Image reads and temporary writes are infrequent and fall below the reporting precision. |
Shared Infrastructure
No shared infrastructure is attributable to DWaste's operational boundary. The benchmark was executed on the iPhone 13 reference device, a dedicated edge device, not a shared cloud environment. No shared-infrastructure allocation is therefore required for E.
Functional Unit
Unit
1 waste-sorting inference (one forward pass on a single image input)
Total During Period
1 functional unit (Benchmark single-inference run)
Justification
A single inference represents the fundamental unit of utility delivered by the software converting an image into a structured 7-class waste classification.
Counting Method
Logged forward-pass executions isolated with OSSignposter during a single-inference benchmark on iPhone 13.
SCI Calculation
SCI = (E × I) + M per R = (5.145 × 10⁻⁷ kWh × 418 gCO2eq/kWh) + 0.000187 gCO2eq = 0.000402 gCO2eq per inference
Energy (E)
Total Energy
5.15 × 10⁻⁷ kWh (derived on iPhone 13, edge reference hardware)
Power Usage Effectiveness (PUE)
N/A (Direct edge runtime deployment)
| Component | Energy | Method |
|---|---|---|
| YOLOv8n-CBAM Edge Inference | 5.15 × 10⁻⁷ | iPhone 13 (MLML3LL/A), iOS 26.5.2; t_inference via OSSignposter; Power Profiler overall-system power 33.7 %/hr at 93% brightness (unplugged) |
Carbon Intensity (I)
Value
418 gCO2eq/kWh
Location
US-PJM (PJM Interconnection)
Approach
Location-based
Data Source
Electricity Maps (historical daily data), 2026
Embodied Emissions (M)
Total Allocated
0.000187 gCO2eq
Methodology
SCI time-share allocation (M = TE × TS × RS), using Apple's published lifecycle data for the reference device.
| Component | Allocated | Source |
|---|---|---|
| TE (Total Embodied, Production + Transport + End-of-life share) | 53,120 gCO2e | Apple iPhone 13 Product Environmental Report (September 2021), 83% of 64 kgCO2e |
| TiR (Time Reserved) | 0.4429 s | Inference duration via OSSignposter |
| EL (Expected Lifespan) | 1.261 × 10⁸ s | 4 years (Apple's stated first-owner use assumption) |
Methodology
Approach
Hybrid measurement on the edge reference device. Inference duration measured directly via OSSignposter in Xcode Instruments; energy derived from battery-percent drain and battery capacity rather than measured at component level; power via Power Profiler overall-system battery drain.
Description
The score is a single-inference benchmark result on iPhone 13, not typical or average deployment performance. The YOLOv8n-CBAM model (3.5 MB ONNX) ran once per inference; the inference interval (442.90 ms) was isolated with OSSignposter. Power Profiler reported overall-system power of 33.7 %/hr at 93% brightness; multiplied by the iPhone 13 battery capacity (12.41 Wh) this gives P_avg = 4.182 W. E = 4.182 W × 0.4429 s ÷ 3,600,000 = 5.15 × 10⁻⁷ kWh. This is an upper bound for isolated inference energy because the power figure is a whole-session average from a separate trace and includes display and system overhead.
Assumptions and Limitations
| Assumption or Limitation | Justification or Mitigation |
|---|---|
| Single-inference benchmark evaluation | Explicitly labelled as a single-inference benchmark on iPhone 13, not typical/average deployment. Benchmark settings do not capture long-term thermal throttling or extended field deployment dynamics; future disclosures will report longitudinal field logs. |
| Power measured at 93% display brightness; overall-system value | Includes display and system overhead, so E is an upper bound for isolated inference energy. Measurement was unplugged on the battery. |
| Embodied emissions allocation | Allocated per SCI time-share using Apple's published LCA and 4-year use assumption; formula and sources in Section 6. |
Self-Certification Statement
This certificate confirms that DWaste has self-certified conformity with ISO/IEC 21031:2024 (Software Carbon Intensity) and provided complete disclosure information verified by the Green Software Foundation.
- Self-certification: DWaste declares that this SCI calculation conforms to ISO/IEC 21031:2024 requirements
- Disclosure verification: GSF verified that all required disclosure elements were provided with sufficient detail for community review
This is not third-party certification or accredited conformity assessment. The disclosed information enables peer review and community validation of conformity claims.
View Full DisclosureReady to Build Sustainable Software?
Take the next step in your green software journey. Get certified, access exclusive resources, and join thousands of practitioners making a real impact on our planet's future.
