Research, indexed.

TFNet introduces a multi-task segmentation architecture with a Dilated ResNet encoder and dual decoders that jointly predict footprint masks and edge maps. The dual heads act as a "tuning fork" that co-regularizes shape and boundary, achieving 94% F1 on SpaceNet2 and WHU benchmarks with strong generalization across sensors.

We quantify how urban parks and green infrastructure mitigate land surface temperature during heatwaves. Combining Landsat LST, Sentinel-2 NDVI, and municipal boundary data, we derive per-park cooling indices and identify high-leverage interventions for climate adaptation planning.

A GCN-based consistency-guided bottleneck that discovers semantically meaningful facial landmarks without supervision. The clustering objective enforces structural consistency across transforms, yielding landmarks competitive with supervised baselines on CelebA and AFLW.

Multi-sensor flood mapping combining Sentinel-1 SAR, Sentinel-2, and Landsat-9 on Google Earth Engine. Applied to the 2022 Pakistan floods, the pipeline identified 1,410 km of damaged road infrastructure with high recall.

PD-SEG disaggregates coarse census data into 30-meter population density maps using building-footprint priors and deep segmentation. Outperforms WorldPop and Meta HRSL on held-out districts, with particularly strong gains in informal settlements.

Cities around the world are facing tremendous pressure due to rapid urbanization. They are being extended haphazardly especially in developing countries, putting strain on already depleting natural recourses. The land-use conversion to built-up areas harms the urban environment significantly. The most immediate implications of this land-use/land cover (LULC) conversion are the transition of Land Surface Temperature (LST) and the creation of Urban Heat Islands. This research investigates the spatial distribution of LST and LULC and their interrelation using satellite images from Landsat 5 (TM) and 8 (OLI/TRS) for the years 1998, 2010, and 2021. The built-up area in Lahore has grown enormously over the last two decades. Our results indicate that each year 1.26% of the land is being transformed into built-up area. Consequently, the prevailing urban development trends have also influenced the LST. In particular, we observed an average upsurge of 0.47°C per year between 1998 and 2021. If our cities continue to expanqd in the same manner, this would have serious ecological implications in the future. Thus, urban planners and policymakers need to incorporate climate-adaptive design at the community and building levels to improve the situation.

Multi-year satellite analysis quantifying urban sprawl trajectories at the city scale. The pipeline fuses segmentation outputs with change-detection to produce longitudinal sprawl indices that correlate with municipal growth indicators.