# Rethinking Streaming-Perception Evaluation on Heterogeneous Edge Platforms

arXiv: https://arxiv.org/abs/2610.05578 · HTML: https://mlsanity.com/paper/2610.05578/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 29 references that are in the map.
- First posted on arXiv 4 Oct 2026.

## Builds on (most cited first, 19 of 19)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Scalability in Perception for Autonomous Driving: Waymo Open Dataset](https://mlsanity.com/paper/1912.04838/) (2019)
- [Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization](https://mlsanity.com/paper/1911.08731/) (2019)
- [Argoverse: 3D Tracking and Forecasting with Rich Maps](https://mlsanity.com/paper/1911.02620/) (2019)
- [TinyLlama: An Open-Source Small Language Model](https://mlsanity.com/paper/2401.02385/) (2024)
- [YOLOv10: Real-Time End-to-End Object Detection](https://mlsanity.com/paper/2405.14458/) (2024)
- [YOLOv12: Attention-Centric Real-Time Object Detectors](https://mlsanity.com/paper/2502.12524/) (2025)
- [Device Placement Optimization with Reinforcement Learning](https://mlsanity.com/paper/1706.04972/) (2017)
- [Real-time Object Detection for Streaming Perception](https://mlsanity.com/paper/2203.12338/) (2022)
- [Are We Ready for Vision-Centric Driving Streaming Perception? The ASAP Benchmark](https://mlsanity.com/paper/2212.08914/) (2022)
- [iGniter: Interference-Aware GPU Resource Provisioning for Predictable DNN Inference in the Cloud](https://mlsanity.com/paper/2211.01713/) (2022)
- [DAMO-StreamNet: Optimizing Streaming Perception in Autonomous Driving](https://mlsanity.com/paper/2303.17144/) (2023)
- [LongShortNet: Exploring Temporal and Semantic Features Fusion in Streaming Perception](https://mlsanity.com/paper/2210.15518/) (2022)
- [StreamYOLO: Real-time Object Detection for Streaming Perception](https://mlsanity.com/paper/2207.10433/) (2022)
- [TOD: Transprecise Object Detection to Maximise Real-Time Accuracy on the Edge](https://mlsanity.com/paper/2105.08668/) (2021)
- [Transtreaming: Adaptive Delay-aware Transformer for Real-time Streaming Perception](https://mlsanity.com/paper/2409.06584/) (2024)
- [Towards Latency-Aware 3D Streaming Perception for Autonomous Driving](https://mlsanity.com/paper/2504.19115/) (2025)
- [BiSwift: Bandwidth Orchestrator for Multi-Stream Video Analytics on Edge](https://mlsanity.com/paper/2312.15740/) (2023)

## Abstract (arXiv)

Multi-camera streaming perception is increasingly deployed on heterogeneous edge platforms shared with co-resident workloads, yet accelerator placement is often evaluated using isolated single-stream experiments and mean streaming average precision (sAP). Using two end-to-end pipelines on a single GPU--NPU platform, we show that isolated evaluation can mis-rank deployment-time placement. Although the GPU pipeline is preferred in isolation, GPU-localized contention introduces deadline misses that make detections stale and can reverse the preferred placement before full GPU saturation. The NPU pipeline is less accurate than the GPU pipeline on small and medium objects in isolation, but nearly matches it on large objects. The largest absolute sAP losses in our latency and contention experiments occur for large objects. In our four-stream experiments, the preferred placement depends on which path becomes stale, and increasing GPU-side contention shifts the best placement from All-GPU to All-NPU. Under a GPU-saturating vision--language co-tenant, All-NPU achieves 5.2× the worst-stream sAP of All-GPU. Because mean sAP can hide severe single-stream degradation, evaluation should report contention sweeps, deadline-miss rates on both paths, and worst-stream sAP alongside mean sAP.