Srijan: Center of Excelance on Generative AI
☕ GenAI for Mobility, Medical and Education ☕
Research Threads
Some Central Theme Line Goes Here
Watermarking in Synthetic Media
A small description goes here
Dehazing and Deraining
A small description goes here
GenAI for HealthCare
Small description goes here
AAAI 2025
Paper ‘TSGAN: Temporal Social Graph Attention Network for Aggressive Behaviour Forecasting’ by Swapnil, Suman and Rajesh acepted for presentation
- First model to forecast individual aggression in social networks.
- TSGAN integrates social influence (via active peers) and temporal decay (via exponential attention) into a unified attention module.
- 24.8% improvement over baselines in daily aggression forecasting on X. State-of-the-art results on cross-platform Flickr for user popularity prediction.
- Proposed LLM-based aggression content detection model (92.87% F1).
- Efficiently scales to moderately large networks and is tolerant to aggression content detection noise.
- Provides actionable forecasts for platforms to mitigate aggression-driven harms (e.g., cyberbullying, offline violence).
WACV 2025
Paper ‘CaMN: Cross-aligned Fusion For Multimodal Understanding’ by Abhishek, Shubham and Suman is acepted for presentation
- A novel denoising based multi-modal architecture that seamlessly integrates vision, text, and graph representation through a proposed guided cross attention mechanism
- Designed a new objective function to align modalities within a unified semantic space, thereby improving the model’s coherence for better classification.
- Superior performance on the publicly available dataset, outperforming existing state-of-the-art methodologies
Checkout our latest publications
Feb 1, 2025
In an era where online aggression increasingly spills into real-world harm—from cyberbullying to offline violence—the ability to predict aggressive behavior before it escalates has become a societal imperative. TSGAN (Temporal Social Graph Attention Network), an AI model that doesn’t just detect aggression but forecasts it, giving platforms and policymakers crucial lead time to intervene.
Feb 1, 2025
Jan 29, 2025
Jan 1, 2025
Jan 1, 2025