We have 8 ZedX cameras that we hope to synchronize and run depth computations on. At the moment we process each camera frames sequentially with pyzed.sl.Camera(...).retrieve_image(...) and pyzed.sl.Camera(...).retrieve_measure(<depth matrix>,sl.MEASURE.DEPTH, ...) in a loop. We are curious how much performance we would gain by batching the camera frames so that inference is only run once. I can’t find a way to do this in the sdk - it seems like retrieve_measure(...) can only be called by one camera at a time.
Hi @neil,
Indeed, with ZED SDK 5.1, image capture and depth processing can only be done per-camera. We would recommend using multi-threading to get the best performance, and running the inferences in parallel. Please take a look at our official Python Multi Camera Sample.
We are preparing a unified API to retrieve all of the synced depth results at once, which has shown some latency and CPU usage improvements for a future release of the ZED SDK.