Abstract: A computer-implemented method including: obtaining a pre-trained upstream machine learning model; fine-tuning the pre-trained upstream model for at least two downstream tasks, wherein the fine-tuning comprises: (a) training a target downstream model for a target downstream task, based on a dataset with labeling specific to the target downstream task, and (b) training a nuisance downstream model for a nuisance downstream task, based on a dataset with labeling of characteristics which are specific to the nuisance downstream task and are undesired for the target downstream task; and normalizing the undesired characteristics from the pre-trained upstream model, to prevent biasing of the target downstream model by the undesired characteristics.

Abstract: A traffic sign recognition system including a detection mechanism adapted for detecting a candidate traffic sign and a recognition mechanism adapted for recognizing the candidate traffic sign as being an electronic traffic sign. A partitioning mechanism may be adapted for partitioning the image frames into a first partition and a second partition. The detection mechanism may use the first portion of the image frames and the recognition mechanism may use the second portion of the image frames. When the candidate traffic sign is detected as an electronic traffic sign, the recognition mechanism may use both the first partition of the image frames and the second portion of the image frames.

Abstract: A traffic sign recognition system including a detection mechanism adapted for detecting a candidate traffic sign and a recognition mechanism adapted for recognizing the candidate traffic sign as being an electronic traffic sign. A partitioning mechanism may be adapted for partitioning the image frames into a first partition and a second partition. The detection mechanism may use the first portion of the image frames and the recognition mechanism may use the second portion of the image frames. When the candidate traffic sign is detected as an electronic traffic sign, the recognition mechanism may use both the first partition of the image frames and the second portion of the image frames.