We compile technical notes you can read with a focus on time-series AI, vision AI, lightweight models, and edge deployment.
All posts are published sequentially. For collaboration or news inquiries, please write to egnox@edgenomix.com.
We have organized how we reflect the context of ISO 13374, ISO 17359, ISO 13381-1, ISO/IEC 23053, ISO/IEC 5259, ISO/IEC TS 6254, and ISO/IEC TR 24029-1 into product design.
Rather than stopping at paper summaries, we have organized what to look at first in the context of the actual floor and operations.
We summarized what to look at when comparing object detection models in complex industrial environments, from a field deployment perspective.
We organized what advantages self-supervised approaches have in environments where labels are scarce.
We explain why lightweight models that can run even on edge devices are needed, in the context of remaining useful life prediction.
We examine how to keep a model lightweight while still securing performance in environments with little data.
We read through a time-series forecasting architecture applicable across various domains from the perspective of industrial data.
We organized the points to watch when connecting a time-series anomaly detection model to actual operational alarms.
We introduce in what situations a label-free unsupervised approach is effective for surface defect detection.
We organized a representative time-series anomaly detection architecture together with industrial field deployment scenarios.
Whether it can be applied depends on the equipment configuration and operational objectives. We can define the scope you need together during the consultation stage.