Structured Differential Learning for Automatic Threshold Setting
This addresses the challenge of efficiently tuning complex rule-based systems in computer vision, though it is incremental as it builds on existing structured approaches.
The paper tackles the problem of automatically tuning parameters in rule-based computer vision systems, such as thresholds and time constants, by introducing a heuristic technique that uses labeled data for approximate gradient descent. The method successfully adjusts over 100 parameters in an automotive headlight controller, outperforming hand tuning with just desired outputs from videos.
We introduce a technique that can automatically tune the parameters of a rule-based computer vision system comprised of thresholds, combinational logic, and time constants. This lets us retain the flexibility and perspicacity of a conventionally structured system while allowing us to perform approximate gradient descent using labeled data. While this is only a heuristic procedure, as far as we are aware there is no other efficient technique for tuning such systems. We describe the components of the system and the associated supervised learning mechanism. We also demonstrate the utility of the algorithm by comparing its performance versus hand tuning for an automotive headlight controller. Despite having over 100 parameters, the method is able to profitably adjust the system values given just the desired output for a number of videos.