NEURAL depth mode execution time

Hello everyone,

I recently started using the NEURAL depth mode with the ZED camera for the first time and observed something intriguing. The initial run took approximately 3 to 4 minutes to complete, which was significantly longer than expected. However, subsequent uses of the NEURAL depth mode were much faster, taking only about 3 to 4 seconds to complete each time.

Can I expect the NEURAL depth mode to consistently perform at this faster speed in future uses, or are there certain conditions or initializations that could cause the initial longer processing time? Iā€™m curious about any insights or explanations for this change in performance time. Thanks


The optimization of the AI model takes a few minutes only the first time you run it. From now it will be fast, all the time, unless you change your ZED SDK or CUDA version.

1 Like