Page 529 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

innovative approach enhances disaster prevention and mitigation capabilities of urban and
community environments, protecting vital economic assets and production spaces.

2�3 Future Work 4.5: Manufacturing

Continuous Improvement

Ongoing data collection and model training will be conducted to ensure the system adapts to
more diverse environmental conditions, maintaining the product's cutting-edge capabilities.

Productization

Based on existing solutions, a standardized product package will be developed, including
hardware (e.g., cameras) and software (AI algorithms and cloud deployment solutions). This will
enable rapid deployment in other ports or logistics centers, reducing implementation barriers.

Platform Development: A cloud-based service platform will be established, integrating visual
large models and other auxiliary tools to provide users with on-demand AI capabilities. The
platform will support a multi-tenant model, allowing customized services according to different
requirements.
Collaboration

Explore collaboration opportunities with other industry partners, such as insurance companies,
to innovate and develop new business models.

Standards Development

Actively participate in the development of relevant national and global standards e.g., in SG21,
sharing our expertise and best practices in the field of AI surveillance for safety management.

3 Use Case Requirements

REQ-01: It is critical that hybrid edge-cloud architecture is required to enable real-time detection
(edge) and contextual validation (cloud), minimizing latency and cloud resource usage.

REQ-02: It is of added value that Edge-optimized lightweight models are required to run
efficiently on low-power devices while maintaining detection accuracy.

REQ-03: It is critical that fine-tuning for both cloud large model and edge small model is
required to improve validation accuracy and adapt to evolving environmental conditions‐

REQ-04: It is critical that unified management interface is required to centralize monitoring of
personnel, vehicles, compliance, and safety inspections across all port zones.

REQ-05: It is expected that dynamic resource allocation is required to balance edge/cloud
workloads based on risk severity and operational priorities.

REQ-06: It is critical that data security protocols are required to encrypt video streams and
operational data across edge/cloud layers.

493