Monitoring COVID-19 social distancing with person detection and tracking via fine-tuned YOLO v3 and Deepsort techniques (2005.01385v4)

Abstract: The rampant coronavirus disease 2019 (COVID-19) has brought global crisis with its deadly spread to more than 180 countries, and about 3,519,901 confirmed cases along with 247,630 deaths globally as on May 4, 2020. The absence of any active therapeutic agents and the lack of immunity against COVID-19 increases the vulnerability of the population. Since there are no vaccines available, social distancing is the only feasible approach to fight against this pandemic. Motivated by this notion, this article proposes a deep learning based framework for automating the task of monitoring social distancing using surveillance video. The proposed framework utilizes the YOLO v3 object detection model to segregate humans from the background and Deepsort approach to track the identified people with the help of bounding boxes and assigned IDs. The results of the YOLO v3 model are further compared with other popular state-of-the-art models, e.g. faster region-based CNN (convolution neural network) and single shot detector (SSD) in terms of mean average precision (mAP), frames per second (FPS) and loss values defined by object classification and localization. Later, the pairwise vectorized L2 norm is computed based on the three-dimensional feature space obtained by using the centroid coordinates and dimensions of the bounding box. The violation index term is proposed to quantize the non adoption of social distancing protocol. From the experimental analysis, it is observed that the YOLO v3 with Deepsort tracking scheme displayed best results with balanced mAP and FPS score to monitor the social distancing in real-time.

• The paper presents a novel framework for real-time COVID-19 social distancing monitoring by integrating YOLO v3 detection with DeepSORT tracking.
• It demonstrates a balanced trade-off between high mean average precision and processing speed, enabling effective surveillance in dynamic environments.
• The proposed method introduces a 'violation index' to quantify social distancing breaches, paving the way for automated public health monitoring.

Monitoring COVID-19 Social Distancing using YOLO v3 and DeepSORT

The article in question addresses a pertinent issue during the COVID-19 pandemic: the enforcement of social distancing in public spaces through technological means. With the rapid spread of COVID-19 and the lack of an immediate medical solution, social distancing has emerged as a critical strategy to mitigate transmission. In response to this need, the authors propose a deep learning framework utilizing real-time surveillance systems for monitoring social distancing practices.

The proposed framework leverages the YOLO v3 (You Only Look Once version 3) object detection model and DeepSORT tracking algorithm to detect and track individuals in video footage. This integration offers a solution capable of segmenting humans from the background with high accuracy and then monitoring their movements to enforce social distancing guidelines. The authors compare the performance of YOLO v3 against other popular state-of-the-art models such as Faster R-CNN and Single Shot Detector (SSD), assessing metrics like mean average precision (mAP), frames per second (FPS), and loss values related to object classification and localization.

Key Methodology and Findings

1. Model Utilization: The authors employ YOLO v3 which deals with real-time object detection as a regression problem rather than a classification problem. The model is augmented with DeepSORT for comprehensive tracking by assigning unique identifiers to detected individuals, facilitating consistent tracking across video frames.
2. Performance Metrics: YOLO v3 with DeepSORT demonstrated superior performance regarding a balanced trade-off between accuracy (mAP) and processing speed (FPS). This balance is crucial for real-time applications where latency could hinder deployment effectiveness.
3. Social Distancing Metrics: The framework applies a vectorized L2 norm computation to determine proximity violations. An innovative 'violation index' is introduced, quantifying the degree of social distancing adherence by measuring the ratio of individuals to identified social groups.
4. Experimental Analysis: The framework's strength lies in its ability to deliver reliable, real-time monitoring with high precision and moderate computational resources. This makes it particularly applicable to environments requiring vigilant enforcement of public health measures.

Implications and Future Directions

This research holds significant practical implications. By automating social distancing monitoring, organizations can more effectively manage compliance in workplaces, public gatherings, and other high-density areas without necessitating constant human oversight. From a theoretical perspective, this paper paves the way for enhancements in autonomous monitoring systems using deep learning, suggesting potential for further research into optimizing detection models for public health and safety applications.

Looking ahead, advancements could focus on refining detection algorithms for complex environments, improving accuracy under occlusion, and integrating additional contextual data, such as environmental factors, which might influence social distancing behavior. Moreover, the adoption of multi-camera setups could enhance tracking accuracy across expansive areas.

In conclusion, the proposed framework provides a robust tool for enforcing social distancing through technological means, demonstrating how AI can play a critical role in managing public health crises. As surveillance technology evolves, integrating such systems with standard safety protocols has the potential to significantly contribute to future pandemic response strategies.