Fix DeepStream Counting
Budget: ₹600 – ₹1,500 INR
I have a Python-based counting application built on NVIDIA DeepStream. It relies on nvsort for tracking, but the tracker IDs keep changing while the same box-shaped objects move through the indoor scene, so the final count is always off.
I will share the current pipeline and counting script. Your job is to inspect the code, identify why nvsort keeps re-assigning IDs, and supply a clean fix that keeps each object’s ID stable from first detection to exit.
What I expect from you:
• corrected or replacement tracker configuration/code that eliminates the frequent ID swaps
• brief notes highlighting the exact drawback you found and how you resolved it
• unchanged counting logic everywhere else unless a modification is essential for reliable results
You should already be comfortable debugging DeepStream (GStreamer pipelines, nvtracker config tuning, deepstream-python bindings) so we can move quickly once I send over the repository.
I will share the current pipeline and counting script. Your job is to inspect the code, identify why nvsort keeps re-assigning IDs, and supply a clean fix that keeps each object’s ID stable from first detection to exit.
What I expect from you:
• corrected or replacement tracker configuration/code that eliminates the frequent ID swaps
• brief notes highlighting the exact drawback you found and how you resolved it
• unchanged counting logic everywhere else unless a modification is essential for reliable results
You should already be comfortable debugging DeepStream (GStreamer pipelines, nvtracker config tuning, deepstream-python bindings) so we can move quickly once I send over the repository.