Department of Computer Science
Permanent URI for this communityhttps://repository.cuilahore.edu.pk/handle/123456789/16
Browse
10 results
Search Results
Item Deep Learning-Based Prediction of Urban Area Expansion(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) Hamza Ali; FA20-RCS-003; LHR TP 8360; Dr. Zeeshan GillaniMulti-temporal data can be useful in many research areas. Multi-temporal data provides us with high-resolution satellite imagery. This high-resolution data can be used to understand change detection in a specific area or the layout of that complete area. To use multi-temporal data, especially custom datasets for deep learning models is real headache. Moreover, Multi temporal data can be very useful with Artificial Intelligence to develop models to tackle different problems like wildfire detection, traffic flow detection, etc. This thesis focuses on two problems 1st one is how can custom multi-temporal dataset which consists of a small sample be used to develop a deep learning model for semantic segmentation purpose of given custom dataset and 2nd problem is related to Urban expansion. The problem of urban expansion is a major issue all over the world, especially in the countries of Africa. Urban expansion has a direct impact on both economic growth and climate change. In this era Machine learning and Computer Vision techniques will provide a vital role to create a model which will help to tackle this problem using multi-temporal data. This thesis methodology is to use custom satellite images data set of a specific area and create a deep learning segmentation model to do segmentation and prediction of different areas in satellite images especially, the urban part. This research focuses on the RGB multi-temporal data set of Dakar, Senegal which is one of the Seaports on the Western Coast of Africa. This thesis methodology is divided into two separate parts. 1st part is about segmentation, an experiment was conducted using with simple Multi U-Net architecture and it achieved more than 85% accuracy on the validation dataset. After these predictions were made using random images from the test dataset. 2nd part focused on the urban expansion problem and trained a Regression model using an Artificial Neural Network which only got a 12% MSE value with 97.25% accuracy and predicts urban expansion on basis of pixel values from remote sensing data using population data from the year 2022 to 2031Item Enhancing Soil Moisture Estimation through Integration of Remote Sensing and Machine Learning Techniques(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2023) Mehwish Mukhtar; SP22-RCS-013; LHR TP 8683; Dr. Zeeshan GillaniSoil moisture (SM) dynamics are integral to effective irrigation management, influencing both crop yield and the conservation of water resources within agricultural practices. Soil Organic matter is a pivotal factor in detecting the soil moisture. The capability to accurately detect irrigation events is essential for water use optimization, a critical concern in regions facing water scarcity where judicious irrigation can bolster crop health and advance sustainable farming practices. Traditional monitoring techniques, reliant on manual field surveys and intermittent data collection, are impeded by their limited scale and lack of continuous temporal coverage. Prior research has predominantly leaned on discrete point-based measurements or episodic remote sensing data, often inadequate for the detailed, persistent monitoring that precision irrigation scheduling demands. Addressing these limitations, our study introduces a cohesive framework for enhanced irrigation detection by using the organic matter of the bare soil, over extensive corn fields by integrating high-frequency satellite imagery with advanced machine learning techniques. Our comprehensive methodology encompasses the acquisition of Sentinel-1 satellite, SMAP and SMOS data, validation with ground-truth references, and sophisticated preprocessing tactics to navigate the common temporal and spatial discrepancies encountered in remote sensing analysis. We assessed various machine learning models, notably. Linear Regression, Random Forest, SVR, and KNN, with Liner Regression demonstrating the most accurate performance in predicting irrigation events and along with other models Naïve Bayes model were used for the Bare Soil classification and Soil Organic matter Detection. Mean Square Error (MSE) and Mean Absolute Error (MAE) are the evaluation parameters for the irrigation detection and F1-Score, Precision, Recall, Support and Accuracy are the evaluation parameters for the Soil Moisture estimation. However, satellite data dependency may introduce atmospheric inconsistencies, and model adaptability across different terrains or crop types may necessitate recalibration. Future directions broader environmental parameters, exploring deep learning potentials, and model validation across diverse agricultural settings. Ultimately, our vision is to develop a universally adaptable platform that delivers real time, actionable insights for stakeholders, revolutionizing irrigation practices on a global scale Soil moisture (SM) dynamics are integral to effective irrigation management, influencing both crop yield and the conservation of water resources within agricultural practices. Soil Organic matter is a pivotal factor in detecting the soil moisture. The capability to accurately detect irrigation events is essential for water use optimization, a critical concern in regions facing water scarcity where judicious irrigation can bolster crop health and advance sustainable farming practices. Traditional monitoring techniques, reliant on manual field surveys and intermittent data collection, are impeded by their limited scale and lack of continuous temporal coverage. Prior research has predominantly leaned on discrete point-based measurements or episodic remote sensing data, often inadequate for the detailed, persistent monitoring that precision irrigation scheduling demands. Addressing these limitations, our study introduces a cohesive framework for enhanced irrigation detection by using the organic matter of the bare soil, over extensive corn fields by integrating high-frequency satellite imagery with advanced machine learning techniques. Our comprehensive methodology encompasses the acquisition of Sentinel-1 satellite, SMAP and SMOS data, validation with ground-truth references, and sophisticated preprocessing tactics to navigate the common temporal and spatial discrepancies encountered in remote sensing analysis. We assessed various machine learning models, notably. Linear Regression, Random Forest, SVR, and KNN, with Liner Regression demonstrating the most accurate performance in predicting irrigation events and along with other models Naïve Bayes model were used for the Bare Soil classification and Soil Organic matter Detection. Mean Square Error (MSE) and Mean Absolute Error (MAE) are the evaluation parameters for the irrigation detection and F1-Score, Precision, Recall, Support and Accuracy are the evaluation parameters for the Soil Moisture estimation. However, satellite data dependency may introduce atmospheric inconsistencies, and model adaptability across different terrains or crop types may necessitate recalibration. Future directions broader environmental parameters, exploring deep learning potentials, and model validation across diverse agricultural settings. Ultimately, our vision is to develop a universally adaptable platform that delivers real time, actionable insights for stakeholders, revolutionizing irrigation practices on a global scaleItem Identification of Best Sowing Patterns of Crop using Satellite Imagery and Deep Learning(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) Zil E Huma Shahzadi; FA19-RCS-020; LHR TP 8337; Dr. Zeeshan GillaniRapid growths in population and climate change are resulting in food insecurity. Food security is more challenged in developing countries like Pakistan where conventional practices of crop monitoring are in place. The two aspects of crop monitoring which are crop identification and yield prediction are manual which are time-consuming and resource extensive thus resulting in delayed decision making. However, the Precision Agriculture applications for crops monitoring based on Remote Sensing data can help us to effectively monitor crops and increase production and aid decision making. This study acquires remotely sensed multispectral and multitemporal satellite imagery from Sentinel-2 mission satellites to extract spectral bands over different timestamps of the rice, wheat, and sugarcane crop season. These spectral bands along with the computed vegetation indices over the growing season of wheat and rice are used with the Long Short Term Memory network for the early identification of the crops. The study also attempted to identify the best sowing dates of wheat crops with multispectral and multitemporal data. Further, the significance of the temporal data and different combinations of the spectral bands were analyzed for crop identification of the small-sized fields and a comparison was made with the existing state of the art. The best combination of the spectral bands resulted in 99.76 % accuracy for crop identification. Further, the crops are also identified with 93.77 % accuracy within the first four weeks of their seeding. Our study also identified the sowing week of wheat crop with RMSE of 0.8. The crop identification accuracy with the proposed approach is suggestive of the applicability of the study for the automatic identification of crops on large scaleItem Integrating Remote Sensing and Deep Learning Techniques for Improved Crop Yield Estimation(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2023) Umer Usman; SP22-RCS-023; LHR TP 8684; Dr. Zeeshan GillaniAccurate yield estimation is crucial for making knowledgeable choices about agriculture and food safety around the world. However, many current studies in this place face issues inclusive of neglecting soil properties as predictors, limited generalization throughout special regions, and confined use of climate variables. This study targets to fill those gaps by using the machine and deep learning and techniques to predict wheat yield in 36 districts of Punjab, Pakistan. Using a dataset created with Google Earth Engine, the dataset consists of detailed statistics on soil residences, climate and other relevant agricultural variables or flora indices. The study included an in-intensity evaluation of the dataset, followed by using training and assessment of various machine learning and deep learning techniques, which includes Artificial Neural Networks, RandomForest, GradientBoosting, and XGBoost.The findings revealed that the ANN model encountered problems together with overfitting because of limited data, while GradientBoosting and XGBoost excel in accuracy and robustness. Specifically, GradientBoosting gave an RMSE of 362.4946 and a MAE of one 101.2008 tons/district, whilst XGBoost gave an RMSE of 280.4313 and an MAE of 102.5128 tons/district. Both algorithms confirmed high R² values, indicating their effectiveness for predicting district-level productiveness. The performance of the models was evaluated with the aid of comparing their predictions with the ground reality information provided by using CRS of Punjab, with error metrics given in tons at district level. The insights gained from this study are very essential to farmers, agronomists and policy makers, allowing them to make informed decisions and promote sustainable agricultural practices. This research makes a big contribution to the sphere of yield estimation and highlights factors for future studies, such as expanding the dataset and exploring other modelling techniques including LSTM, to enhance protection of food and sustainable agricultureItem City Scale Multi Camera Vehicle Tracking Using Deep Learning(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2023) Muhammad Sameed Khan; SP20-RCS-013; LHR TP 8465; Dr. Zeeshan GillaniIntelligent city and traffic management depend on city-scale multi-camera vehicle tracking; however, this work has various difficulties. On the different viewing angles, problems arise that include the variation on the large scale, frequent occlusion and appearance variation. In this study, we use the cross-camera tracking technique and multi-camera tracking system that considers aggregation loss. To address the challenges of multi-camera vehicle tracking, the suggested system has four key parts. First, we extract the tracks with the help of a single camera view by identifying the object and multi-object tracking modules. These modules combine their detection capabilities to provide efficient tracking between frames. After obtaining the tracklets, we use a multi camera re-identification module to match them. The tracklets acquired by several cameras are connected by this module using re-identification techniques. We used OsNetX1_0 and retinaNet50 for reidentification. In the final stage, we deal with the isolated trackless and tracking of the synchronize ids that rely on the outcomes of the re-identification. We improve the computational performance with the help of less parameter models and cohesion of the tracking system by removing isolated tracklets and assuring consistent tracking IDs. With practical and effective alternatives, the research advances multi-camera vehicle tracking on the city scale. However, this system shows the significance of the fast-multi-target cross-camera tracking approaches and the loss of aggregation in the study challenges. The AI city challenges' success reveals our system's potency and viability.Item Multisource Remote Sensing Land Cover Classification for Crop Identification and Yield Prediction using Multispectral Temporal Imagery(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Haseeb Rehman Khan; FA19-RCS-017; LHR TP 7593; Dr. Zeeshan GillaniRapid growths in population and climate change are resulting in food insecurity. Food security is more challenged in developing countries like Pakistan where conventional practices of crop monitoring are in place. The two aspects of crop monitoring which are crop identification and yield prediction are manual which are time-consuming and resource extensive thus resulting in delayed decision making. However, the Precision Agriculture applications for crops monitoring based on Remote Sensing data can help us to effectively monitor crops and increase production and aid decision making. This study acquires remotely sensed multispectral and multitemporal satellite imagery from Sentinel-2 mission satellites to extract spectral bands over different timestamps of the rice, wheat and sugarcane crop season. These spectral bands along with the computed vegetation indices over the growing season of wheat and rice are used with the Long Short Term Memory network for the early identification of the crops. The study also attempted to predict the yield of wheat crops with multispectral and multitemporal data. Further, the significance of the temporal data and different combinations of the spectral bands were analyzed for crop identification of the small-sized fields and a comparison was made with the existing state of the art. The best combination of the spectral bands resulted in 99.76 % accuracy for crop identification. Further, the crops are also identified with 93.77 % accuracy within the first four weeks of their seeding. The Root Means Square Error of 2. 61 and 71.67 % R-squared is reported for the prediction of the wheat yield. The crop identification accuracy with the proposed approach is suggestive of the applicability of the study for the automatic identification of crops on large scale.Multisource Remote SensingItem Metropolis Violation Detector Using Deep Neural Network(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Iqra Khalid; FA19-RCS-014; LHR TP 7298; Dr. Zeeshan GillaniAnomaly discovery on road traffic is a significant assignment because of its extraordinary potential in metropolitan traffic the board and street wellbeing. It's anything but an extremely difficult assignment since the strange occasion happens seldom and shows various practices. In this work, we present a model to recognize abnormality in street traffic by gaining from the vehicle movement designs in two particular yet connected modes, i.e., the static mode and the powerful mode, of the vehicles. The static mode investigation of the vehicles is gained from the foundation demonstrating followed by vehicle identification technique to find the unusual vehicles that keep still out and about. The unique mode investigation of the vehicles is gained from identified and followed vehicle directions to find the strange direction which is atypical from the predominant movement designs. The outcomes from the double mode investigations are finally melded by driven a re identification model to acquire the final abnormality. This study was based on three classes of anomalies (car crash, car stall and lane change). It was divided into a two class problem with 63 anomaly folders and 86 no anomaly folders. Dataset was provided by NVIDIA AI city challenge track 4 containing 100 train and 100 test videos. Firstly videos were annotated according to anomaly event time. The anomaly time was converted into seconds and then specific frames were cropped. Then 30 frames per video were selected to fed into ResNet18 for high feature extraction. Then LSTM architecture and 3D-CNN was trained and results were evaluated on accuracy evaluation measure. Satisfying results was obtained from both architectures i-e; 80 % accuracy of 3D-CNN and 83 % accuracy of ResNet with LSTM. As it is an on-growing field many work can be done in future by using advanced models for the improvementsItem Computational Technique For Classification Of Membrane Proteins(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2020) Harmain Asghar; SP18-RCS-012; LHR TP 5985; Dr. Zeeshan GillaniThe Cell is the most important part of the human body, many small entities and molecules are encapsulated within a cell; shielded by the cell membrane. Proteins are the basic elements within the cell that are responsible for various functionalities, like nutrients across membrane and molecule, identification of foreign bodies, etc. Membrane proteins play a very important function in predicting diseases. Due to advancements in the Next-generation Sequencing, membrane proteins dataset is ever increasing. The experimental methods for the classification of these proteins are very expensive and time-consuming. In this study, prediction about the membrane proteins and its subtypes will be made based on a computational approach using protein primary sequence and machine learning algorithm. For the purpose of more improvement we used the convolutional neural network inception v3 model. The results exhibit the effectiveness of predictive performance as compared to the other existing algorithm. Some useful techniques are described to make the study more effective.Item Pipeline for Identification of Terrorist Organization and Classification of Their Members Using Social Media(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2020) Sana kousar; FA17-RCS-023; LHR TP 5978; Dr. Zeeshan GillaniSocial media has a tendency to change perception or opinion of people. The surge in use of social media has made people venerable to exploitation by different banned outfits. The aim of this research is to analyze social media content and to identify potential individual or group of individual and their target. We generated a dataset using Twint API and use natural language processing techniques to preprocess the data and later use network analysis and different machine learning ensemble techniques to identify potential individuals that are targeting and recruiting people in the name of NGO’s.Item Stock Market Prediction Using Deep Learning(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Hamza Javaid; CIIT/SP22-RCS-014/LHR; Dr. Zeeshan Gillani; LHR TP 9696Accurate stock market prediction remains a complex research challenge due to the market’s irregular, non-linear, and highly dynamic multivariate nature. Traditional statistical methods often struggle to capture the volatile patterns, and deep dependencies present in stock time-series data. This thesis addresses the problem by proposing a hybrid deep learning-based fusion model designed to improve the accuracy of stock price trend forecasting. The proposed solution adopts a late fusion approach, integrating the complementary strengths of four models: Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) to capture sequential dependencies, the Temporal Fusion Transformer (TFT) for modeling complex temporal relationships through attention mechanisms, and the Multi-Layer Perceptron (MLP) for nonlinear feature interactions in time-series data. This architecture forms a robust forecasting system, trained on multivariate time-series data from the NASDAQ-100, which includes both raw stock metrics and derived technical indicators. The fusion model achieved an R² score of 0.99196 and an MSE of 0.00033, clearly outperforming standalone LSTM, GRU, and stacked LSTM-GRU baselines. The predictions closely follow actual market movements with minimal lag, capturing both bullish and bearish trends effectively. This research presents a high-accuracy predictive framework that offers meaningful contributions to the field of stock forecasting, helping investors make informed buy/sell/hold decisions, thereby reducing risk and improving investment strategies.