Camera parameter enhancement for multiple analytics
Methods and systems for camera configuration include configuring an image capture configuration parameter of a camera according to a multi-objective reinforcement learning aggregated reward function. Respective quality estimates for analytics are determined after configuring the image capture parameters. The aggregated reward function is updated based on the quality estimates.
1 . A computer-implemented method for camera configuration, comprising:
configuring an image capture configuration parameter of a camera according to a multi-objective reinforcement learning aggregated reward function;
determining respective quality estimates by a plurality of analytics unit quality estimators after configuring the image capture parameters; and
updating the aggregated reward function based on the quality estimates from the plurality of analytics unit quality estimators.
2 . The method of claim 1 , wherein determining the quality estimates includes applying respective trained estimator models that have been trained to accuracy for respective analytics tasks.
3 . The method of claim 1 , wherein updating the aggregated reward function combines the quality estimates according to an aggregation strategy.
4 . The method of claim 3 , wherein the multi-objective reinforcement learning uses a linear aggregation strategy.
5 . The method of claim 3 , wherein the multi-objective reinforcement learning uses a winner-takes-all aggregation strategy.
6 . The method of claim 3 , wherein the multi-objective reinforcement learning uses a weighted aggregation strategy.
7 . The method of claim 1 , wherein the image capture configuration parameter is selected from the group consisting of control brightness, contrast, color, sharpness, and focus.
8 . The method of claim 1 , further comprising training respective quality estimation models, for the plurality of analytics, to determine quality estimates based on an input image.
9 . The method of claim 1 , further comprising capturing a new image with the camera after changing the image capture configuration parameter, wherein determining the quality estimates is done using the new image.
10 . The method of claim 1 , wherein the multi-objective reinforcement learning treats a present set of image capture configuration parameters as a state of the camera and uses the aggregated reward function to determine an action that reflects a change in one or more of the image capture configuration parameters to balance performance of the plurality of analytics.
11 . A system for camera configuration, comprising:
a hardware processor; and
a memory that stores a computer program which, when executed by the hardware processor, causes the hardware processor to:
configure an image capture configuration parameter of a camera according to a multi-objective reinforcement learning aggregated reward function;
determine respective quality estimates by a plurality of analytics unit quality estimators after configuring the image capture parameters; and
update the aggregated reward function based on the quality estimates from the plurality of analytics unit quality estimators.
12 . The system of claim 11 , wherein the computer program further causes the hardware processor to apply respective trained estimator models that have been trained to accuracy for respective analytics tasks.
13 . The system of claim 11 , wherein the computer program further causes the hardware processor to combine the quality estimates according to an aggregation strategy.
14 . The system of claim 13 , wherein the multi-objective reinforcement learning uses a linear aggregation strategy.
15 . The system of claim 13 , wherein the multi-objective reinforcement learning uses a winner-takes-all aggregation strategy.
16 . The system of claim 13 , wherein the multi-objective reinforcement learning uses a weighted aggregation strategy.
17 . The system of claim 11 , wherein the image capture configuration parameter is selected from the group consisting of control brightness, contrast, color, sharpness, and focus.
18 . The system of claim 11 , further comprising training respective quality estimation models, for the plurality of analytics, to determine quality estimates based on an input image.
19 . The system of claim 11 , wherein the computer program further causes the hardware processor to capture a new image with the camera after changing the image capture configuration parameter, wherein determining the quality estimates is done using the new image.
20 . The system of claim 11 , wherein the multi-objective reinforcement learning treats a present set of image capture configuration parameters as a state of the camera and uses the aggregated reward function to determine an action that reflects a change in one or more of the image capture configuration parameters to balance performance of the plurality of analytics.