Bharat in Betastartup field guide
← Back to the atlas
No. 1310Problem statementSpace and research

Hybrid AI–NWP Multi-Model Forecast Blending System

Source

Source summary

The hybrid AI–NWP framework should adaptively weight physical models, ensembles and AI forecasts according to historical skill, region, season, lead time and weather regime. Outputs should cover rain, temperature, wind and extreme-weather guidance, with model-weight maps and an automated operational workflow. Improved skill relative to individual models is an evaluation objective, not a proven result.

No. 1046Problem statementSpace and research

Target detection by optimizing Anomaly Detection in Hyperspectral Image Processing using AI/ML

Research AI/ML anomaly detection in public hyperspectral datasets such as Hyperion or AVIRIS, identifying spectrally unusual pixels and object signatures. The source discusses preprocessing, denoising, fusion and radiometric/atmospheric correction before evaluating deep-learning models. This overview retains image-analysis research scope without target-engagement instructions; improved detection remains to be demonstrated.

No. 1112Problem statementSpace and research

Integrated Geo-Referenced Fish Catch Data Repository and Access System

Create an INCOIS repository consolidating georeferenced fish-catch records from researchers, industry, cruises and image-based collection. Accept Excel data and organise by date, location, depth and species, with filters, visualisation and downloads for species-specific habitat research. Preserve distinctions between species-level weights and aggregate catch weights with species presence. Use separate occurrence and abundance modules, geographic/time filters, mapped results, Excel/CSV exports and authenticated access with an updating administration interface.

No. 1114Problem statementSpace and research

Data download Duplication Alert System (DDAS)

Develop a dataset-download duplication alert for institutes and research organisations. Maintain download metadata and use history or unique identifiers to recognise a dataset even under another filename. Before downloading, notify users of an existing copy in institutional repositories or user accounts, including its location, timestamp, period and spatial coverage. The source envisages applicability across research, academia, government and other sectors.