In an embodiment, a camera sensor module is communicatively coupled to a vision processing system. The camera sensor module captures a plurality of exposures, produces an HDR-combined stream, and sends both the HDR-combined stream and the plurality of exposures to the vision processing system. A first ISP of the vision processing system performs color processing to produce a color-processed HDR-combined stream for neural network processing, presentation to a human, and/or computer vision processing. At least one other ISP of the vision processing system performs color processing to collect exposure-specific statistics on at least one of the plurality of exposures for determining camera calibration data for the camera sensor module.