Search for: All records

Vision–language models learn visual concepts from the supervision of natural language. It can significantly enhance the generalizability of real-time intelligent sensing, such as analyzing camera-captured real-time images for visually impaired users. However, adapting vision–language models to distribution shifts at test time, caused by several factors such as lighting or weather changes, remains challenging. In particular, most existing test-time adaptation methods rely on gradient-based fine-tuning and backpropagation, making them computationally expensive and unsuitable for real-time applications. To address this challenge, the Training-Free Dynamic Adapter (TDA) has recently been introduced as a lightweight alternative that uses a dynamic key–value cache and pseudo-label refinement for test-time adaptation without backpropagation. Building on this, we propose TDA-L, a new framework that integrates Low-Rank Adaptation (LoRA) to reduce the size of feature representations and related computational overhead at test time using pre-learned low-rank matrices. TDA-L applies LoRA transformations to both query and cached features during inference, cost-efficiently improving robustness to distribution shifts while maintaining the training-free nature of TDA. Experimental results on seven benchmarks show that TDA-L maintains accuracy but achieves lower latency, less memory consumption, and higher throughput, making it well-suited for AI-based real-time sensing. 

Real-time object detection is essential for AI-based intelligent traffic management. However, growing complexities of deep learning models for object detection cause increased latency and resource requirements. To tackle the challenge, we introduce a new approach, named AROD (Adaptive Real-Time Object Detection), that infers the pixel motion speed in continuous traffic video frames and skips redundant frames when the pixel velocity is low. Thereby, AROD aims to significantly enhance the efficiency and scalability, sustaining the accuracy of object detection. Our evaluation using real-world traffic videos reveals that our method for pixel velocity inference via lightweight deep learning reduces the RMSE (Root Mean Square Error) by up to two orders of magnitude compared to state-of-the-art approaches. AROD improves the frame processing rate of YOLOv5, SSD, and EfficientDet by approximately 32-61\%, 110-174\%, and 120-213\%, respectively. AROD considerably enhances scalability by supporting real-time object detection for up to three concurrent traffic video streams on a commodity machine. Moreover, AROD demonstrates its generalizability by supporting competitive accuracy in object detection for a separate traffic video that was fully hidden during training. 

Real-time object detection is essential for AI-based intelligent traffic management. However, growing complexities of deep learning models for object detection cause increased latency and resource requirements. To tackle the challenge, we introduce a new approach, named AROD (Adaptive Real-Time Object Detection), that infers the pixel motion speed in continuous traffic video frames and skips redundant frames when the pixel velocity is low. Thereby, AROD aims to significantly enhance the efficiency and scalability, sustaining the accuracy of object detection. Our evaluation using real-world traffic videos reveals that our method for pixel velocity inference via lightweight deep learning reduces the RMSE (Root Mean Square Error) by up to two orders of magnitude compared to state-of-the-art approaches. AROD improves the frame processing rate of YOLOv5, SSD, and EfficientDet by approximately 32-61%, 110-174%, and 120-213%, respectively. AROD considerably enhances scalability by supporting real-time object detection for up to three concurrent traffic video streams on a commodity machine. Moreover, AROD demonstrates its generalizability by supporting competitive accuracy in object detection for a separate traffic video that was fully hidden during training. 

Intelligent mobile image sensing powered by deep learning analyzes images captured by cameras from mobile devices, such as smartphones or smartwatches. It supports numerous mobile applications, such as image classification, face recognition, and camera scene detection. Unfortunately, mobile devices often lack the resources necessary for deep learning, leading to increased inference latency and rapid battery consumption. Moreover, the inference accuracy may decline over time due to potential data drift. To address these issues, we introduce a new cost-efficient framework, called Corun, designed to simultaneously handle multiple inference queries and continual model retraining/fine-tuning of a pre-trained model on a single commodity GPU in an edge server to significantly improve the inference throughput, upholding the inference accuracy. The scheduling method of Corun undertakes offline profiling to find the maximum number of concurrent inferences that can be executed along with a retraining job on a single GPU without incurring an out-of-memory error or significantly increasing the latency. Our evaluation verifies the cost-effectiveness of Corun. The inference throughput provided by Corun scales with the number of concurrent inference queries. However, the latency of inference queries and the length of a retraining epoch increase at substantially lower rates. By concurrently processing multiple inference and retraining tasks on one GPU instead of using a separate GPU for each task, Corun could reduce the number of GPUs and cost required to deploy mobile image sensing applications based on deep learning at the edge. 

Deep learning models have significantly improved object detection, which is essential for visual sensing. However, their increasing complexity results in higher latency and resource consumption, making real-time object detection challenging. In order to address the challenge, we propose a new lightweight filtering method called L-filter to predict empty video frames that include no object of interest (e.g., vehicles) with high accuracy via hybrid time series analysis. L-filter drops those frames deemed empty and conducts object detection for nonempty frames only, significantly enhancing the frame processing rate and scalability of real-time object detection. Our evaluation demonstrates that L-filter improves the frame processing rate by 31–47% for a single traffic video stream compared to three standalone state-of-the-art object detection models without L-filter. Additionally, L-filter significantly enhances scalability; it can process up to six concurrent video streams in one commodity GPU, supporting over 57 fps per stream, by working alongside the fastest object detection model among the three models.