技術摘要(英)
This technology is applied to an already mature AI model, which is currently deployable in three target environments (Taiwan and India) for street scene scenarios. In the future, if new applications arise, such as (1) different weather conditions, (2) changes in camera field of view (FOV), (3) alterations in camera installation positions, or (4) expansion to new regions that require re-collection of training datasets, this technology can be leveraged to reduce the number of labeled samples needed for retraining.