Batch processing multiple camera streams in parallel

The DeepStream SDK processes 30 streams in parallel. I want to perform something similar where I have 15 camera streams and 3 NVIDIA 1080ti for performing Deep Learning operations on those streams. All my models have been written in Tensorflow and have been converted to TensorRt format. Need to understand the best possible way to handle the 15 streams in parallel and perform batch inference on them. I prefer not using the DeepStream SDK for now and need some alternative solution using open source components


Will recommend to use DeepStream for batch processing multiple camera streams in parallel:
Please refer below link for more details: