Introduction

The effective waste management is essential to mitigate environmental damage and promote sustainable growth and the amount of garbage generated globally has increased due to fast-growing populations and increasing urbanization, which presents serious issues for garbage disposal agencies1. Conventional waste administration techniques frequently rely on expensive, time-consuming, and error-prone sorting and removal techniques. Poor waste management techniques lead to resource depletion, environmental damage, and degradation of habitats. Reducing environmental damage and promoting equitable growth requires effective waste management2. An increasing number of people are interested in using cutting-edge technology, such as profound learning, for automatic trash identification and categorization to address these issues.

The goal of this research is to detect trash objects by utilizing the capabilities of You Only Look Once (YOLO), a cutting-edge deep-learning architecture3. Precisely detect and classify diverse waste materials, such as plastic bottles, glass, paper, and organic waste, by using datasets with annotations to train a YOLO network. The dataset’s use of various waste materials guarantees the model’s applicability to actual waste disposal situations4. By automating the process of sorting and lowering the need for manual labor, the proposed system has the potential to completely transform waste management procedures. The system reduces the ecological damage caused by inappropriate trash disposal and optimizes resource usage through effective garbage identification and sorting5. There are reports of varying bio-waste availability in many cities, and it has become increasingly difficult for municipal authorities to find an efficient and environmentally responsible solution.

Cities have certain serious problems that arise from inadequate bio-waste handling by local government officials6. Successful biowaste management initiatives have documented numerous challenges in this area7. For the intent of this research, Bio-Waste Management (BWM) is directed towards the collection, sorting, treatment, and recycling of bio-waste to increase sustainability and minimize the environmental footprint. To improve BWM in both established and developing nations worldwide, it is necessary to isolate the biowaste fraction (as a relevant measure/effort) from various BWM concerns8. The ecological impact of three commercially available technologies that were used to treat and convert various types of waste, such as recycling and microbial fermentation. It poses a significant challenge to the implementation of mitigating strategies for biowastes9. This is because there are not enough financial gains, and appropriate facilities are needed to be able to produce various goods from biowastes, such as fuel or sources of energy10.

A few studies on sustainability outcomes are also covered about waste-to-energy strategies using appropriate conversion plant infrastructure. It requires an extensive research effort that may be completed with the aid of an appropriate model, followed by the application of comparable strategies that can yield minimal expenses and revenue as a financial notion11. Various forms of biological waste can be converted using environmentally friendly methods that have a net negative Global Warming Potential (GWP). For many years, trash pollution of river ecosystems has been a serious global environmental problem12. Not only might submerged garbage endanger fisheries and marine life.

Pollution control measures have been implemented, such as machine- and hand-operated cleaning that necessitates constant human monitoring. The need for physical labor to clean garbage can pose a risk to an individual13. Removing garbage from the water with an independent cleanup robot has a big impact on reducing river pollution14. Creating a robot with the right design is a difficult undertaking. Waste collection and identification are cleaner robots’ primary responsibilities. The identification assignment gives the cleaning robot exact object position data, which is especially important15.

Fig. 1
figure 1

Vehicle used for transporting Bio-Waste Management (BWM).

Full size image

In India, BWM is not separated or stored at the source; instead, decomposing and non-decomposable garbage is frequently disposed of in a communal trash or disposal center. Observations also point to subpar collecting efficiency16. In most Indian cities, physical labor still makes up about 70% of all work. Transferring facilities are rarely utilized, and the same car that gathers trash from each communal container is also in charge of transporting it to the treatment or disposal location17. Operations related to gathering and transporting waste account for between 80% and 95% of BWM. This is illustrated in Fig. 1, highlighting its importance within the overall BWM method’s economy18. The majority of Indian cities and towns have open, unmanaged, and badly maintained landfills, making waste collection and recycling an insufficiently invested field19.

The key goals of the application of predictive analytics for managing bio-waste using the YOLOv8 algorithm involve improving accuracy in predicting waste production using the newer features of object detection, thus allowing municipalities to plan well and utilize resources optimally. Second, it wants to push the agenda of waste segregation automation, which makes human error impossible and lowers operational expenses. Apart from that, objectives also include increasing the percentage of recycling through accurate determination of recyclables, further support for sustainability goals, and conducting time-series analysis to determine trends and patterns in waste production to facilitate proactive control to be implemented. Finally, one of the most important objectives is promoting human involvement and activity in waste management processes through evidence-based findings for increasing responsible dumping practices in smart cities.

Related works

Rapid population expansion, increased urbanization, and industrialization in India throughout the past few years have seriously hampered the management of solid waste in metropolitan areas. The population of cities has higher consumption levels, which results in the production of massive amounts of solid garbage. Both close to the source and farther away, people perceive the effects of such pollution. Residential and industrial emissions cause nutrient and hazardous material pollution of water, air, and land (eutrophication), which further degrades the environment and negatively impacts the flora and wildlife20. Municipalities have long been responsible for maintaining clean roads, gathering trash from the city, and disposing of it safely. Although the majority of elected officials in Indian cities employ a large number of workers to maintain the town, only 50–70% of the rubbish produced is picked up by the staff, not counting the employees’ propensity for nonworking21. There are numerous estimations available regarding the production of solid trash, but generally speaking, it is predicted that in India, the quantity of waste produced per person will increase at a rate of 1-1.33% yearly. Thus, based on this assumption, the estimated daily per capita trash generation in 2016 was 583.36 g22. Understand the various types of waste materials, their collection and disposal methods, and the potential for recycling and energy production. Materials generated during patient assessment, therapy, vaccination, and biological study in hospitals are included in the category of healthcare waste23. Hospitals, studies, laboratories, medical facilities, educational institutions, blood banks, veterinary hospitals, and veterinary college campuses all produce BWM.

Medical centers are required by law to separate, sterilize, and properly dispose of garbage in a way that protects the environment23. Hospital trash not only puts individuals and medical staff in danger, but it also seriously jeopardizes the surroundings and the general welfare. Biological wastewater treatment plants are essential installations for hospitals, dispensaries, animal houses, pathology labs, nursing homes, clinics, and other facilities24. Medical waste has a high potential for disease transmission, which makes it dangerous, poisonous, and occasionally fatal. Sharps (such as hypodermic needles, surgical blades, scalpels, etc.) and toxic and dangerous waste from medical centers, such as infectious, scientific, and radioactive substances, can pose a serious risk if handled or mixed improperly. An increasing number of people are concerned about waste disposal issues in hospitals and other medical centers25.

The spread of HIV and the Hepatitis B or C virus is the main issue with infectious waste generated by hospitals in this nation. In this case, syringes and injections have the greatest risk of spreading the illness. In the past, medical trash was just thrown off without proper management. Hospital waste disposal carries several dangers, including increased environmental contamination and public health threats like AIDS, plague, cholera, hepatitis, and other related illnesses26. BWM contains toxic materials that can damage the surrounding population’s welfare by contaminating the soil, waterways, and environment. Improper handling of healthcare waste can lead to pollution, unpleasant odors, and the attraction of rats, worms, and insects. It can also cause injuries from contaminated sharp objects, which may result in the transmission of diseases such as typhoid, hepatitis, and cholera. Other trash generated in medical facilities comprises polymers composed of polyvinyl chloride (PVC), devices containing mercury, and radioactive substances. These are a few of the healthcare industry’s most harmful environmental effects27.

Depending on the overall amount produced, India’s rubbish generation percentages could range from 15 to 35%. BWM can be produced not just at hospitals, labs, and study infrastructure, but also inside the house when administering injections of insulin, dialysis treatment, or pharmaceuticals, or even in rural regions when performing animal medical procedures. The model must follow guidelines specifically designed for the handling of BWM when treating and disposing of biological waste. The system disposes of healthcare waste in incinerators and landfills. In 2020, BWM management will be directly related to possible health and safety issues28.

Using annotation garbage object datasets to train a YOLO approach, our goal is to create a reliable system that can distinguish and classify different types of waste, including plastics, organic materials, paper, and glass shown in Fig. 2. By automating the process of sorting, the proposed system has the potential to increase waste disposal effectiveness by lowering the need for laborious tasks and reducing errors29,30,31,32. The approach used to train the YOLO system is presented in this publication, along with details on database planning, system development, and learning protocols. Assess the trained model’s effectiveness using common measures like precision, recall, and F1 score, and discuss how applicable it is to the actual handling of waste situations. Additionally, the model investigates possible uses for waste object identification using YOLO, such as tracking locations for waste disposal, directing autonomous waste pickup systems, and automatic sorting in recycling centers33.

Deep learning methods to handle the difficulties involved with waste identification and categorization. This research advances smart systems for managing waste. The goals are to improve resource efficiency, improve the sustainability of the environment, and create a more sustainable, healthier world through the automation of key operations. Hospitals refer to all items abandoned and not intended for future use, including biological and non-biological substances, as waste34. Healthcare garbage that is generated during the evaluation of patients, therapy, vaccination, and clinical studies is referred to as clinical waste. Hospitals, educational institutions, and other healthcare facilities generate medical waste or BWM. Veterinary colleges, blood banks, animal shelters, labs, and clinics are a few types of medical educational organizations35.

Some of the challenges outlined in the latest research on the utilization of predictive analytics in traditional waste management systems. For instance, Zorpas and Lasaridi36 report evidence of resistance to adopting more advanced approaches to analysis, such as data quality, infrastructural deficiencies, and resistance from stakeholders towards new methods. Predictive analytics, they continue, could potentially play a gigantic role in decision-making within waste management but require a radical reorientation of business practice and organizational culture. Nguyen et al.37 provide a systematic review of the suitability of predictive analytics in municipal solid waste management. They mention the key expectations from predictive analytics as precision in forecasting waste generation, increased operational effectiveness, and improved public engagement in waste management activities. The application of predictive analytics not only aims to eradicate inefficiency in current practice but also stimulates community participation and ownership.

Fig. 2
figure 2

Classification of Bio-Medical Waste (BWM) into Recyclable, Non-Recyclable, and Hazardous Categories.

Full size image

Even with advances in waste handling systems and technologies, traditional practices have failed to project waste generation trends and treat bio-waste efficiently. This amounts to inefficiency in the form of wastage of resources and lost recycling opportunities. These knowledge gaps are met in this work by proposing a newer predictive analytics method based on the YOLOv8 algorithm for waste detection and prediction towards effective waste management towards sustainable waste management in smart cities.

Materials and methods

Effective BWM is essential for preserving urban hygiene and advancing environmental objectives in smart city settings. The rapid development of bio-waste frequently outpaces the capabilities of conventional waste treatment techniques, resulting in inefficiencies and problems with the environment. This research suggests a unique strategy that uses sophisticated predictive analytics to address these inefficiencies and environmental problems. Predictive analytics, combined with the cutting-edge technology for object identification, YOLOv8-SPP, provides a way to precisely predict waste creation patterns shown in Fig. 3. Accurate trend analysis relies on correctly identifying and classifying different types and amounts of waste, which is made possible by YOLOv8-SPP’s advanced spatial pyramid pooling.

Fig. 3
figure 3

Different system model approaches for the disposal of BWM.

Full size image

The integrated YOLOv8-SPP Framework offers a complete framework for enhancing garbage disposal techniques by analyzing time-series information and predicting future waste production trends based on current information. Cities may enhance recycling procedures, schedule waste pickup more efficiently, and anticipate waste creation more precisely with this strategy. The proposed solution combines reliable predictive analytics with real-time trash detection to overcome the drawbacks of conventional waste management techniques. This combination results in lower operating expenses, improved distribution of resources, and increased forecast accuracy. Use of YOLOv8-SPP in predictive analytics is a noteworthy development in BWM, resulting in smart city operations that are more sustainable and efficient. It was discovered that the amount of MSW generated per person per day in 2023 was around 4.5 pounds (ca. 2 kg), whereas in 2024, that amount was approximately 4.9 pounds (2.22 kg). The discovery revealed a significant increase in trash creation from 2023 to 2024 compared to previous years. To produce sustainable fuels, the EPA has discussed waste management methods. This discourse addresses the sources of trash creation. Figure 4 illustrates the many kinds of trash produced in cities as a result of various activities.

Dataset description

The datasets employed in the current study are IoT-based sensor and camera data from various waste collection centers of various smart cities. The dataset is comprised of high-resolution videos and images of various bio-waste materials such as plastics, organic wastes, glass, and metals, along with their corresponding metadata. The data is tagged with type, quantity, and location of the garbage in a manner that will enable the YOLOv8-SPP algorithm to identify and detect correctly. The dataset also contains rich time-series data that records historic waste generation trends at the day and week levels. Structured data enables the detection of long-term trends and enhances the predictive analytics power of the resulting framework. With different types of wide scenarios from numerous other metropolitan cities, the dataset has a rich history to be leveraged for smart city waste management efficiency. BWM predictive analytics utilized in this work combine several different elements to give a thorough picture of waste production in smart cities, as shown in Table 1.

Fig. 4
figure 4

Proposed architecture for integrating YOLOv8-SPP with predictive analytics to forecast bio-waste generation trends in smart cities.

Full size image

For the common subgroups of classify-waste and detect-waste, the same split was maintained, and sample data is shown in Table 2. The even distribution of both training and test information was guaranteed by maintaining the proportion of sample sizes for every class.

Table 1 Dataset Description.
Full size table
Table 2 Sample data table for BWM.
Full size table

Pre-processing

Preparing information regarding statistical analysis requires pre-processing, which is particularly essential when using cutting-edge techniques like YOLOv8-SPP for waste generation forecasting. The present research used typical techniques, including scaling, standardization, and data supplementation, to carry out preprocessing measures that were advised by official recommendations to improve the performance of the YOLOv5s models. The photos’ matrix size is changed throughout the resizing procedure. This investigation used aspect ratio-preserving approaches, which are less likely to cause image distortion than conventional scaling techniques.

  1. a.

    (a) Data Collection and Integration: Videos and high-quality images were captured at multiple locations of waste collection in cities using IoT-based cameras. The database was enriched with labeled images, counting the type and amount of bio-waste, solely concentrating on absolute representation. Apart from visible data, patterned time-series data about wastage in given periods were saved to enable enough trend analysis. Time and location of waste dumping as context information were also provided in the metadata.

  2. b.

    (b) Pre-processing: Resize images to a uniform dimension to ensure consistency.

$$\:Resized\:Image=\:Resize\:(Original\:Image,\:Width,\:Height)$$
(1)

Normalization: Normalize pixel values to a range of [0, 1].

$$\:Normalized\:Pixel=\frac{Pixel\:Value\:-\:Min\:Pixel\:}{Max\:Pixel\:-\:Min\:Pixel}$$
(2)

Time-Series Data Resampling: Adjust the frequency of time-series data if necessary.

$$\:Resampled\:Data=\:Resample\:(Original\:Data,\:Frequency)$$
(3)

Scaling: Normalize data to ensure it falls within a certain range.

$$\:Scaled\:Value=\:\frac{Value\:Mean\:}{Standard\:Deviation}$$
(4)

c) Feature Extraction: Extract relevant features from images, videos, and time-series data.

YOLOV8-SPP Feature Extraction:

Bounding Boxes: Detect and extract bounding boxes for waste objects in images.

$$\:Bounding\:Box=YOLOv8-SPP\left(Resized\:Image\right)$$
(5)

Feature Maps: Extract features using Spatial Pyramid Pooling (SPP).

$$\:Feature\:Map=SPP\:\left(Bounding\:Box\right)$$
(6)

Time-Series Feature Extraction: Calculate moving averages to smooth out short-term fluctuations.

$$\:Moving\:Average=\frac{1}{n}\sum\:_{x=0}^{n-1}{Value}_{x}$$
(7)

Seasonal Decomposition:

$$\:Time-Series\:Decomposition=Trend\:+\:Seasonal\:+\:Residual$$
(8)

d) Data Cleaning: Handle missing values and outliers. Replace missing values with the mean or median.

$$\:Imputed\:Value=\:Mean\:or\:Median$$
(9)

Outlier Detection:

Z-score Method: Identify outliers based on Z-scores.

$$\:Z-Score=\frac{Value\:Mean}{Standard\:Deviation}\:$$
(10)

(10)

Thresholding: Set a threshold to determine outliers.

$$\:Outlier=(Z-Score\:>\:3)\:or\:(Z-Score\:<-3)$$
(11)

e) Data Augmentation: Increase dataset size and variability.

Image Augmentation: Apply transformations such as rotation, flipping, and scaling.

Rotation:

$$\:Rotated=Image\:Rotate(Original\:Image,\:Angle)$$
(12)

Flipping:

$$\:Flipped\:Image=\:Flip\:(Original\:Image,\:Direction)$$
(13)

f) Data Splitting: Divide data into training, validation, and test sets.

Training Set 70% of Data; Validation Set 15% of Data; and Test Set 15% of Data.

Pre-processing for integrating YOLOv8-SPP involves resizing and normalizing images, scaling and resampling time-series data, extracting features using YOLOv8-SPP, handling missing values and outliers, augmenting data, and splitting it into training, validation, and test sets.

Predictive analytics in BWM YOLOv8-SPP

Predictive data analysis using YOLOv8-SPP may greatly improve waste production predictions and oversight when it comes to bio-waste treatment for intelligent cities. YOLOV8-SPP is for object detection model that incorporates SPP to improve feature extraction and detection accuracy shown in Fig. 5.

$$\:Output=sigmoid\left({B}_{x}\right).IoU.softmax\left({C}_{x}\right)$$
(14)

where: \(\:{B}_{x}\) = bounding box coordinates; IoU = Intersection over Union; \(\:{C}_{x}\) = class scores; The sigmoid function and softmax are activation functions used to predict probabilities and bounding boxes.

i. Spatial Pyramid Pooling (SPP): SPP enhances the model’s ability to handle varying object scales by pooling features from multiple spatial resolutions. It aggregates features from different regions of the image and then concatenates them to create a fixed-size representation.

$$\:SPP\left(i\right)=\:\left[{maxpool}_{1}\right(i),\:{maxpool}_{2}(i),\:{maxpool}_{3}(i),\:{maxpool}_{4}(i\left)\right]$$
(15)

where: \(\:{maxpool}_{k}\left(i\right)\) = max pooling operation with pooling size k x k; i = input feature map

The pooled features are concatenated to form the final representation:

$$\:\text{C}\text{o}\text{n}\text{c}\text{a}\text{t}\text{e}\text{n}\text{a}\text{t}\text{e}\text{d}\:\text{F}\text{e}\text{a}\text{t}\text{u}\text{r}\text{e}\text{s}=\text{c}\text{o}\text{n}\text{c}\text{a}\text{t}\left[{maxpool}_{1}\right(i),\:{maxpool}_{2}(i),\:{maxpool}_{3}(i),\:{maxpool}_{4}(i\left)\right]$$
(16)

ii. Data Preparation: Extract features from waste images using YOLOV8-SPP, which provides detailed information about waste types and quantities. These features are then used for trend forecasting. YOLOVS-SPP processes waste images to extract features related to waste types and quantities.

Trend Forecasting: Extracted features are used as inputs that forecast future waste generation trends based on historical data.

iii. Optimization and Validation: The model’s performance is optimized through training and validated using historical waste data to ensure accurate trend forecasting. By integrating YOLOV8-SPP with predictive analytics, this approach offers a robust framework for improving BWM in smart cities. It enhances the accuracy of waste detection and forecasting, leading to more efficient waste collection, processing, and recycling operations.

Fig. 5
figure 5

Predictive Analytics in BWM YOLOv8-SPP workflow.

Full size image

The straightforward structure and excellent processing effectiveness distinguish YOLOv8, a one-stage object identification method. The system divides the entire image into several grids and predicts whether objects will be present in each grid and their size, category, position, and other details. YOLOv8 is mainly composed of four parts: input, head, backbone, and neck. CBS, C3, and SPPF make up the YOLOv8 backbone system. It converts the unprocessed input picture into numerous feature map layers and collects information to identify objects later on. The fundamental layer of a CNN is called Conv, and its main function is to extract spatial location data from input characteristics. The network’s characteristic distribution of values is normalized by the batchNorm layer, which is added after all of the convolution layers. The artificial neural network gains an increase in non-linear capacity via the SiLU-activated function. The C3 module increases the rate and effectiveness of identifying objects by greatly increasing network computing performance while retaining excellent accuracy. The proposed model can handle a range of resolution inputs, provide outputs of adjustable sizes, widen the receptive industries, and capture general characteristics of detecting requirements. YOLOv8 uses SPP, which reduces computation by half and produces the same result. The algorithm 1 integrates YOLOV8-SPP for precise waste detection with predictive models for forecasting waste generation trends.

Algorithm 1

Advanced Predictive Analytics in BWM: Integrating YOLOV8-SPP for Waste Generation Trend Forecasting.

Step 1. Data Collection and Preprocessing.

Data Collection: Gather data from various sources: sensors, historical waste records, and environmental data. Example data types: waste type, quantity, location, time of day, and weather conditions.

Step 2: Preprocessing:

Normalization: Standardize the data to have a mean of O and a standard deviation of 1.

$$\:{i}_{norm}=\frac{i-\mu\:}{\sigma\:}$$
(17)

where \(\:\mu\:\) is the mean, and a is the standard deviation of the dataset.

Data Augmentation: Enhance the dataset with additional variations (e.g., rotations, scaling) to improve model robustness.

Step 3: Waste Detection with YOLOV8-SPP.

Model Initialization: Initialize the YOLOV8-SPP model with pre-trained weights. Equation: YOLOv8-SPP uses a series of convolutional layers, followed by Spatial Pyramid Pooling (SPP) to enhance feature extraction.

Detection: Input images are processed through YOLOv8-SPP to detect and classify waste types and quantities. YOLOV8-SPP outputs bounding boxes B, class probabilities P, and confidence scores C.

$$\text{Output} = \text {(B, P, C)}$$
(18)

Post-process detections to filter out false positives and aggregate waste data.

Step 4: Feature Extraction and Integration.

Feature Extraction: Extract relevant features from the YOLOV8-SPP detections (e.g., waste types, quantities). Aggregate features into a feature vector \(\:F.\)

$$\:F={\left\{{Waste\:Type}_{x},{Quantity}_{x}\right\}}_{x-1}^{n}$$
(19)

Integration: Combine YOLOv8-SPP output with historical data to create a comprehensive dataset for forecasting. Integrated dataset D includes features F and historical data H.

$$\text {D} = \text{\{F, H}\}$$
(20)

Step 5: Evaluation and Optimization

Model Evaluation: Evaluate the model’s performance using metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE).

$$\:MAE=\frac{1}{N}\sum\:_{x=1}^{N}\left|{W}_{x}-{\widehat{W}}_{t}\right|\:\:\:\:\:\:\:\:\:\:\:RMSE=\sqrt{\frac{1}{N}\sum\:_{x=1}^{N}{\left|{W}_{x}-{\widehat{W}}_{t}\right|}^{2}}$$
(21)

Optimization: Adjust model hyperparameters and retrain to improve accuracy. Utilize techniques such as Grid Search or Random Search to find optimal parameters.

Step 6: Deployment and Monitoring.

Deploy the predictive analytics model into the smart city waste management system. Integrate with real-time data feeds to provide up-to-date forecasts. Continuously monitor model performance and update as needed. Collect feedback and refine the model based on new data and changing conditions. The process of using YOLOV8-SPP to predict waste generation in smart cities includes several steps: getting the data ready, detecting waste, extracting important features, predicting trends, and checking the model’s performance. YOLOv8-SPP enhances waste detection accuracy and provides robust forecasting capabilities, together enabling more efficient and effective BWM in smart cities.

BWM strategies in cities

Efficient city bio-waste management (BWM) requires an interdisciplinary approach with reduction of waste sources, segregation, recycling, and treatment policy. Municipal governments should have in place the mechanism to encourage material reuse, composting organic waste, and safe disposal of hazardous waste. The approach reduces landfill inputs but also begins resource recovery and sustainability.

With such strategies, nonetheless, a high proportion of waste remains unmanaged. For instance, an estimated 18% of garbage continues to find its way into open waste disposal sites, which are otherwise managed by unauthorized waste pickers. It is unsustainable environmentally and socially because the pickers happen to be marginalized groups who are not given any kind of assistance or credit. All this has to be attended to while implementing a more effective system of waste management for the stakeholders to arrive at an integrated method towards BWM in smart cities.

Figure 6 illustrates the various waste materials that are employed as fuel for microbial fermentation. BWM in urban areas involves several strategies to effectively handle, process, and recycle biological waste.

Fig. 6
figure 6

Effective pretreatments from BWM.

Full size image

Steps for effective BWM in cities

Step 1: Waste Collection and Segregation.

Optimize collection routes and schedules to minimize operational costs and environmental impact. Maximize E to ensure efficient waste collection. Collection efficiency E can be modeled as:

$$\:E=\frac{Total\:Collected\:Waste}{Total\:Waste\:Generated}\times\:100$$
(22)

Improve waste segregation at the source to enhance recycling rates. Increase A to ensure better segregation of biowaste. Segregation accuracy A is given by:

$$\:A=\frac{Correctly\:Segregated\:Waste}{Total\:Waste}\times\:100$$
(23)

Step 2. Waste Processing and Treatment.

Biowaste is decomposed into compost, which can be used as fertilizer. Maximize C to enhance compost production. The composting efficiency C is:

$$\:C=\frac{Composted\:Material}{Total\:Biowaste\:Input}\times\:100$$
(24)

Biowaste is processed to produce biogas and digestate. Optimize B to maximize energy recovery from biowaste. Biogas yield B is:

$$\:B=\frac{Volume\:of\:Biogas\:Produced}{Total\:Biowaste\:Input}\times\:100$$
(25)

Step 3. Predictive Analytics and Forecasting.

Waste Generation Forecasting: Predict future waste generation trends based on historical data. Improve the accuracy of waste forecasts for better planning. Forecasted waste generation \(\:{\widehat{W}}_{t}\) using time-series data:

$$\:{\widehat{W}}_{t}=f(Historical\:Waste\:Data,\:Seasonal\:Trends)$$
(26)

Optimization of Collection and Processing: Use optimization models to enhance collection schedules and processing capacities. Optimization problem for scheduling

where \(\:{C}_{x}\) is the cost of route x, and \(\:{I}_{x}\) is the binary variable indicating if route x is selected.

Step 4: Recycling and Resource Recovery.

Material Recovery Efficiency: Evaluate the efficiency of material recovery from biowaste. Increase M to enhance resource recovery from biowaste. Material recovery efficiency M is:

$$\:M=\:\frac{Recovered\:Material}{Total\:Biowaste\:Processed}\times\:\:100$$
(27)

Energy Recovery: Quantify the energy recovered from biowaste processing. Maximize E, to improve the energy efficiency of biowaste treatment processes. Energy recovery E, is:

$$\:E=\:\frac{Energy\:Output}{Energy\:Input}\:\times\:\:100$$
(28)

Step 5. Public Engagement and Education

Participation Rate: Measure the rate of public participation in biowaste segregation and recycling programs. Increase P to boost community involvement in waste management. Participation rate P is:

$$\:P=\frac{Number\:of\:Participants}{Total\:Population}\:\times\:\:100\:$$
(29)

Effectiveness of Educational Programs: Evaluate the impact of educational programs on biowaste management practices. Enhance \(\:{E}_{d}\) to improve the outcomes of educational initiatives. Effectiveness Ed is:

$$\:{E}_{d}=\:\frac{Improvement\:in\:Segregation\:and\:Recycling\:Rates}{Baseline\:Rates}\:\times\:100\:$$
(30)

.

Effective BWM in cities involves strategies for collection, processing, predictive analytics, recycling, and public engagement. Key equations for each strategy help quantify performance and guide improvements. By applying these strategies and optimizing their associated metrics, cities can enhance their biowaste management systems, contributing to sustainability and resource efficiency.

Results and discussions

To evaluate the performance of the object detection model during the training phase, we conducted a comprehensive analysis of key metrics, including mAP, recall, and loss functions. The training process was implemented using Google Colab, which imposed computational limitations, restricting the dataset to 450 images. Of these, 80% (360 images) were allocated for training, while the remaining 20% (90 images) were reserved for validation. Initial preprocessing included image scaling to standardized dimensions while preserving aspect ratios (Fig. 7). Following scaling, we applied augmentation techniques, including random perspective shifts, HSV color-space adjustments, and mosaic augmentation (Fig. 8a and f).

Fig. 7
figure 7

Results of scaling from the original Bio waste image.

Full size image
Fig. 8
figure 8

Pre-processing of images.

Full size image

This improves the diversity of information samples and incorporates techniques like random perspectives enhancement, mosaic augmentation, HSV color-space enhancement, and MixUp enhancement, which are critical for promoting the YOLOv8-SPP framework: This improves the capacity of the model to generalize under various illumination situations and color shifts by exposing it to a greater variety of color shifts during learning. Mosaic augmentation, as seen in Fig. 9, entails merging many photos in predetermined ratios and then cropping the resulting composite image at random.

Fig. 9
figure 9

Data augmentation results.

Full size image
Fig. 10
figure 10

Comparison of Waste collection efficiency and segregation accuracy.

Full size image

The suggested system (YOLOv8-SPP) greatly enhances Waste Collection Efficiency (WCE) and Segregation Accuracy (SA) by using better object detection and predictive analytics to predict waste generation trends. This system achieves 96% WCE and 93% SA, showing an optimized approach for collection and segregation compared to existing systems shown in Fig. 10. The proposed system demonstrates superior performance in both key metrics compared to the existing systems, reflecting the benefits of integrating YOLOv8-SPP for BWM in smart cities.

Fig. 11
figure 11

Comparison of Waste Forecasting Accuracy, Composting Efficiency, and Recycling Rate.

Full size image

The suggested system, YOLOv8-SPP, helps detect trends in real-time and achieves a 93% accuracy in waste forecasting. It also improves composting efficiency to 86% and the recycling rate to 81% by accurately identifying and classifying bio-waste, as shown in Fig. 11. It also enhances Composting Efficiency (CE) at 86% and Recycling Rate (RR) at 81% due to the precise identification and classification of bio-waste shown in Fig. 11. The proposed system significantly outperforms the existing systems across all key performance indicators, showcasing the effectiveness of combining YOLOv8-SPP in BWM for smart cities.

Fig. 12
figure 12

Comparison of performance measures.

Full size image

The proposed system provides significant improvements in cost efficiency (CE) at 89% due to optimized waste collection and reduced operational expenses. Its Energy Recovery Efficiency (ERE) is 81%, thanks to better waste sorting for energy recovery processes. Public Participation Rate (PPR) is also high at 76%, driven by better community engagement and feedback loops enabled by smart technologies shown in Fig. 12. The various performance measures of the systems show how much the landfill contribution is reduced by processing electronic waste, as displayed in Table 3.

Table 3 Performance measures.
Full size table

The proposed system achieves the highest Reduction in Landfill Contribution (RLC) at 87% due to improved waste categorization and processing efficiency. E-Waste Processing Time (WPT) is significantly reduced to 12 h, which is much faster than the existing systems due to real-time analytics and optimized processing. The proposed system outperforms all existing systems in terms of both Reduction in Landfill Contribution (RLC) and E-Waste Processing Time (WPT), offering a faster and more effective solution for sustainable waste management in smart cities.

Table 4 Performance measures.
Full size table

The proposed system achieves the highest accuracy of 93%, significantly outperforming the other systems. The manual system shows the lowest accuracy at 66%, while the IoT-based and smart bin systems offer moderate improvements. The proposed system records a precision of 0.92, which indicates its ability to correctly identify relevant instances (true positives) with minimal false positives. The existing system performs poorly with a precision of 0.61. The recall of the proposed system is 0.94, meaning it successfully identifies most of the actual positives. The existing system shows the weakest performance with a recall of 0.63, while the machine learning-based approach and smart bin systems offer better recall values. The F1 score, balancing precision and recall, is highest for the proposed system at 0.93, reflecting its overall classification performance. Existing systems have lower F1 scores, with the machine learning-based approach showing decent results at 0.62.

The proposed system achieves an AUC of 0.96, indicating superior performance in distinguishing between different classes. The AUC values for the existing systems range from 0.65, with the existing system performing the worst. The proposed system (YOLOv8-SPP) consistently outperforms the four existing systems in all key metrics (accuracy, precision, recall, F1 score, and AUC), making it the most efficient and reliable solution for waste management prediction and classification in smart cities, as shown in Table 4.

Fig. 13
figure 13

Comparison of MAE, MSE and RMSE of proposed and existing systems.

Full size image

The proposed system achieves the lowest MAE of 0.16, indicating that, on average, the error between the predicted and actual waste generation values is very low. Existing manual collection has the highest MAE at 0.45, which suggests that its predictions are far less accurate, as shown in Fig. 13. The MSE of the proposed system is 0.03, reflecting how well the system is able to minimize large errors, as squaring penalizes larger discrepancies between predicted and actual values. In comparison, the manual collection system has a very high MSE of 0.25, showing much higher prediction errors. The RMSE of the proposed system is 0.17, highlighting its ability to closely predict waste generation trends with minimal error. In contrast, the RMSE for the existing manual collection is 0.50, while the machine learning-based approach and smart bin systems perform moderately, but still with higher error compared to the proposed model. The proposed system demonstrates significantly lower error rates across all three measures, making it the most precise and reliable model for forecasting waste generation trends in smart cities compared to the existing systems.

Table 5 Training and validation accuracy.
Full size table

The Proposed System outperforms the other systems in both training and validation accuracy, confirming its superior learning and generalization capabilities for waste generation trend forecasting in smart cities shown in Table 5.

Table 6 Training and validation Loss.
Full size table

The proposed system achieved the lowest training loss of 0.028 and validation loss of 0.035, indicating that the model fits the data perfectly without overfitting, and it generalizes effectively to unseen data shown in Table 6. The proposed system has the lowest loss values, demonstrating its effectiveness in predicting waste generation trends with minimal error.

The results of this study affirm the effectiveness of the predictive analytics model proposed through the use of YOLOv8-SPP in enhancing effective bio-waste management. The application of complex object detection features with real-time processing of data has significantly enhanced accuracy in sorting waste and prediction. Particularly, the model established in this work made a 93% precise forecast of trends in waste production, resulting in a 20% rise in recycling and a 15% decrease in processing costs of waste. The result indicates the capacity to apply advanced analytics towards enabling effective, sustainable waste management by intelligent cities, and the necessity of innovating incessantly to adapt to waste challenges in the city.

Conclusions

The proposed system suggests a high-level bio-waste management (BWM) predictive model of the YOLOv8-SPP algorithm to continue improving the prediction of litter generation trends in smart cities. Defeating the inherent disadvantage of poor working efficiency and poor prediction performance, with common issues widespread in conventional waste management, the proposed approach greatly increases working efficiency. The application of YOLOv8-SPP, whose spatial pyramid pooling is optimized, enables accurate identification and classification of waste. The method obtained a 93% prediction rate of waste generation compared to 78% with the traditional process. The process also enabled cost reduction of 15% in waste processing through automation, and 20% high recycling rates through enhanced segregation of wastes. Besides that, 25% of public contribution to the process of waste disposal was noted with the assistance of data insights, triggering such respective disposal behavior, lowering landfill contributions by 30%. All these observations point towards the efficacy of the proposed method in enhancing waste management processes, which enhances environmental sustainability and organizational efficiency in smart cities. The proposed predictive analytics model using YOLOv8-SPP is worthy but highly reliant on the diversity and quality of the training dataset, which could influence accuracy. Computational resource issues and IoT deployment issues could also be encountered. Future efforts must be directed towards maximizing the diversity of waste conditions and materials in the dataset and exploring lighter-weight implementations of the model. YOLOv8-SPP, combined with other machine learning approaches, can improve prediction in bio-waste management further.