AdaMM: Adaptive Object Movement and Motion Tracking in Hierarchical Edge Computing System

Abstract

This paper presents a novel adaptive object movement and motion tracking (AdaMM) framework in a hierarchical edge computing system for achieving GPU memory footprint reduction of deep learning (DL)-based video surveillance services. DL-based object movement and motion tracking requires a significant amount of resources, such as (1) GPU processing power for the inference phase and (2) GPU memory for model loading. Despite the absence of an object in the video, if the DL model is loaded, the GPU memory must be kept allocated for the loaded model. Moreover, in several cases, video surveillance tries to capture events that rarely occur (e.g., abnormal object behaviors); therefore, such standby GPU memory might be easily wasted. To alleviate this problem, the proposed AdaMM framework categorizes the tasks used for the object movement and motion tracking procedure in an increasing order of the required processing and memory resources as task (1) frame difference calculation, task (2) object detection, and task (3) object motion and movement tracking. The proposed framework aims to adaptively release the unnecessary standby object motion and movement tracking model to save GPU memory by utilizing light tasks, such as frame difference calculation and object detection in a hierarchical manner. Consequently, object movement and motion tracking are adaptively triggered if the object is detected within the specified threshold time; otherwise, the GPU memory for the model of task (3) can be released. Moreover, object detection is also adaptively performed if the frame difference over time is greater than the specified threshold. We implemented the proposed AdaMM framework using commercial edge devices by considering a three-tier system, such as the 1st edge node for both tasks (1) and (2), the 2nd edge node for task (3), and the cloud for sending a push alarm. A measurement-based experiment reveals that the proposed framework achieves a maximum GPU memory reduction of 76.8% compared to the baseline system, while requiring a 2680 ms delay for loading the model for object movement and motion tracking.

RAVA: Region-Based Average Video Quality Assessment

“…There do exist some advanced object detection and tracking methods such as AdaMM [ 41 ] and content-aware focal plane selection [ 42 ]. They can handle complex situations such as occlusion.…”

Cognitive Video Surveillance Management in Hierarchical Edge Computing System with Long Short-Term Memory Model

“…Interestingly, ref. [18] revealed that if there was no appearance of an object during a specified threshold time, the heavy DL model could be released to save unnecessary GPU memory consumption. Even though [18] achieved a performance superior to other approaches in terms of GPU memory consumption reduction, an unnecessary delay when reloading the model might happen depending on the corresponding hyperparameter of the threshold time.…”

“…[18] revealed that if there was no appearance of an object during a specified threshold time, the heavy DL model could be released to save unnecessary GPU memory consumption. Even though [18] achieved a performance superior to other approaches in terms of GPU memory consumption reduction, an unnecessary delay when reloading the model might happen depending on the corresponding hyperparameter of the threshold time. For example, if the threshold time value was too large (e.g., 30 s), then the release frequency of the DL model decreased and consumed more GPU memory, whereas if the threshold time value was too small (e.g., 10 s), a DL model release and reload switching frequently occurred, causing reloading delays.…”

“…In that case, when the motion tracking model was loaded for observing motion, the baseline kept holding GPU memory for tracking motion despite the absence of an object in the video frame. -AdaMM: All suggested modules (i.e., frame differences and management of adaptive processes) were included in AdaMM [18].…”

Privacy retrieval method of big data in mobile network based on edge computing

“…On the other hand, most of the existing searchable encryption technologies are based on the design of tight cipher algorithm. The calculation is usually large, and multiple rounds of interaction between users and ECs are required, which greatly increases the communication between users In addition, the existing ciphertext query system model is only suitable for single user system, and only pays attention to the precise query of single keyword However, in practical application, multi-user model and multi keyword similarity search are more common [4]. At present, there is a lack of effective methods for ciphertext query in multi-user model.…”