activeCycle 2026.1
Low-Latency INT8 Quantized Object Detection on Embedded Jetson
Profiling post-training calibration loss and FPS throughput of YOLOv8n on resource-constrained ARM/NVIDIA architectures with layer-wise sensitivity audits.
POD 02
Edge AI & Vision Systems
Designing latency-critical perception pipelines, INT8/FP4 quantization schemes, multi-camera telemetry streaming, and embedded hardware acceleration across NVIDIA Jetson micro-architectures. We study empirical tradeoffs between frame latency, thermal throttling, and precision degradation on resource-constrained platforms.
Core Research Questions
- What is the empirical Pareto frontier between INT8 calibration loss and inference FPS on embedded edge GPUs?
- How can zero-copy CUDA memory layouts prevent frame bottlenecks in multi-stream RTSP ingestion?
- What architectural modifications allow YOLO-family backbones to maintain mAP under aggressive low-bit integer quantization?
Scientific Methodology
Post-Training Quantization (PTQ) & Quantization-Aware Training (QAT)
Hardware-in-the-loop thermal and power telemetry
TensorRT engine layer fusion and kernel autotuning
Zero-copy frame ingestion buffers
proposedCycle 2026.2
Multi-Camera Video Stream Anomaly Detection Pipeline
Designing a zero-copy frame ingestion buffer for temporal anomaly localization in campus perimeter feeds without CPU memory serialization bottleneck.
benchmark report2026-03-25
AIAC-BR-2026-01: Comparative INT8 vs FP16 Inference Latencies for YOLOv8 on NVIDIA Jetson Hardware
Empirical benchmarking of frame throughput, memory bandwidth saturation, and thermal profiling across TensorRT compiled vision graphs on Jetson embedded boards. Reports exact latency curves across batch sizes 1 to 8 under realistic ambient thermal conditions.