Applying Inference over Specific Frame Regions with NVIDIA DeepStream

Originally published at:

This tutorial shares how to apply inference over a predefined area of the incoming video frames.

Great article! I find it very insightful and shows the hidden power of DeepStream SDK.

A small question: After Figure 1 there is a paragraph that starts with " Here are the performance metrics when yolov4 is applied over the entire frame compared to over the tile.", How can we see this comparison?


I have question is it possible to update ROI at runtime using NvDsPreprocessing or any other way?