top of page
Search

Glove Detection for Hygiene Compliance

Introduction

In many industries, particularly in food processing, healthcare, and pharmaceuticals, personal protective equipment such as disposable gloves is essential for maintaining hygiene and safety standards. Failure to wear this protective gear can lead to contamination, health hazards, and violations of regulatory guidelines. However, manually monitoring every individual to ensure compliance is neither efficient nor always reliable.


In this blog, we will discuss hairnet and mask detection and implement code to detect them.



Problem Statement

Regulatory authorities such as the Food and Drug Administration and local health departments require the use of gloves in specific settings to prevent cross contamination and protect public health. However, failure to follow glove protocols, whether due to negligence or oversight, can result in serious hygiene violations, significant fines, and even temporary closures.


Key challenges organizations face include:

  • Inconsistent glove usage by staff during critical activities

  • Limited ability to monitor multiple employees in real time

  • Difficulty in maintaining accurate compliance records

  • Human error and fatigue during manual supervision



How Glove Detection Systems Work

Artificial intelligence based detection systems use computer vision and deep learning to automatically recognize whether individuals are wearing gloves while performing tasks. These systems are powered by advanced machine learning models such as YOLO (You Only Look Once), which are trained on large datasets to accurately distinguish gloved from ungloved hands.


Core capabilities of these systems include:

  • Glove compliance detection: Identifying whether gloves are worn correctly

  • Real time monitoring: Continuously analyzing video feeds to detect non compliance

  • Automated alerts: Sending immediate notifications when gloves are not worn

  • Compliance logging: Generating detailed records for inspections and audits



Benefits of Implementation in Restaurants and Healthcare Settings

1. Improved Hygiene Control: Automated systems enhance cleanliness by ensuring consistent glove use, reducing the chance of contamination.

2. Lower Risk of Violations: Preventing hygiene breaches protects customers and patients, and minimizes the risk of disease transmission or foodborne illness.

3. Efficient Compliance Reporting: AI systems provide complete digital records, simplifying documentation during regulatory inspections.

4. Increased Staff Awareness: Real time feedback promotes good hygiene practices and supports targeted retraining efforts when necessary.

5. Financial Protection: Avoiding penalties, legal actions, or reputational damage due to safety violations can result in significant financial savings.



Applications in Real World Environments

Glove detection powered by artificial intelligence is already being used in various industries:

  • Food service and production: Verifies glove usage in food handling areas

  • Medical facilities: Ensures doctors and nurses wear gloves during procedures

  • Pharmaceutical manufacturing: Maintains sterile conditions in production environments

  • Public health efforts: Supports hygiene monitoring during disease outbreaks



Dataset Used

For this project, we will use two datasets. The PPE 10.3 dataset (https://universe.roboflow.com/ppe103/ppe10.3/dataset/2) contains 9,886 images across all labels, including boots, glasses, face masks, gloves, no gloves, and other items.


The PPE Detection 4 dataset (https://universe.roboflow.com/training-y4lu7/ppe-dataset-4/dataset/1) contains images for boots, glasses, gloves, helmets, and vests, with a total of 8,343 images across all labels.


For this project, we will focus exclusively on glove-related images from both datasets. These datasets will be processed to create a binary classification system with only two labels: "Gloves" and "No Gloves."



Implementation

Dataset Configuration

The foundation of any successful computer vision project lies in proper data organization. Our system uses a structured approach:

dataset_config = {
    'path': abs_dataset_path,
    'train': 'train/images',
    'val': 'valid/images',
    'names': {
        0: 'gloves',
        1: 'no_gloves',
    }
}

This configuration defines two critical classes: workers wearing gloves (compliant) and workers without gloves (violation). The binary classification approach keeps the model focused and highly accurate.


Training Pipeline

The training function incorporates several best practices for production-ready models:

def train_safety_detection_model(
    dataset_yaml_path,
    epochs=120,
    batch_size=16,
    image_size=640,
    model_size='s',
    patience=20
):

Key Training Parameters:

  • 120 epochs ensure thorough learning without overfitting

  • Early stopping (patience=20) prevents wasted computation time

  • YOLOv5s model balances speed and accuracy for real-time applications

  • 640px image size provides detection resolution


Real-Time Video Processing Engine

The heart of the system lies in its video processing capabilities. Here's what makes it powerful:

def detect_safety_equipment_in_video(
    model_path,
    video_path,
    output_path,
    conf_threshold=0.5,
    iou_threshold=0.45
):

Detection Logic:

  • Confidence threshold (0.5) filters out uncertain detections

  • IoU threshold (0.45) eliminates duplicate bounding boxes

  • Real-time FPS calculation monitors system performance

  • Frame-by-frame analysis ensures no violations are missed


Compliance Assessment

The system goes beyond simple detection by providing actionable safety insights:

# Count detections by class
gloves_detected = sum(1 for det in detections if det['class'] == 'gloves')
no_gloves_detected = sum(1 for det in detections if det['class'] == 'no_gloves')

# Determine compliance status
if gloves_detected > 0 and no_gloves_detected == 0:
    status = "COMPLIANT"
    color = (0, 255, 0)  # Green
elif no_gloves_detected > 0:
    status = "VIOLATION DETECTED"
    color = (0, 0, 255)  # Red


Result






Full code is available at the following link:




Insights

Dataset Diversity Critical for Gloves Detection

The initial YOLOv5s model with 2,514 training images failed to detect gloves because the training data and validation data were too similar. However, when the dataset was expanded to 7,603 images by combining multiple datasets, performance improved significantly.


Gloves Detection Suffers from Visual Similarity Issues

Some models consistently confused gloves with other objects, particularly mistaking caps for gloves and detecting forearms as "not wearing gloves." This suggests gloves lack distinctive visual features that differentiate them from other cloth-based objects, requiring more targeted training data.


Annotation Quality Issues Specific to Gloves

The YOLO11s experiment highlighted that combining people and gloves datasets might have created annotation inconsistencies - images with people often lacked glove labels.


Segmentation Approaches Ineffective for Gloves

When transitioning to segmentation models (YOLO-Seg), the performance for hand/gloves detection remained poor despite high precision scores (0.985). The models rarely segmented gloves correctly and often misidentified other objects as hands/gloves.


Gloves Detection Requires Hand Detection Pipeline

The experiments with hand detection models (MediaPipe, DeepLabv3) revealed that these models often failed to detect hands when gloves were present.


For more insights, refer to this link:



Get Help When You Need It

Developing a glove detection system can be challenging, especially when addressing real-time video analysis, achieving high detection accuracy for objects like gloves, or deploying the solution into production environments such as restaurants, hospitals, or pharmaceutical facilities. Don’t hesitate to seek expert support when facing issues related to dataset quality, model training inconsistencies, or video stream processing bottlenecks.


If you’re working on a glove detection system and need personalized assistance, CodersArts offers guidance for both students and enterprises implementing computer vision technologies.


  • For students working on academic or research projects, CodersArts provides help with model configuration, debugging glove detection pipelines, dataset annotation strategies, and evaluating detection performance.

  • For enterprises, we deliver production-grade solutions, including system architecture consultation, real-time compliance monitoring integration, edge deployment for low-latency environments, and customization aligned with industry hygiene standards and operational needs.


Visit www.codersarts.com or email contact@codersarts.com to get the support you need to build, optimize, and scale your glove detection system.




bottom of page