July 18 2026 VISION Transformers - ViTCap-R

Hi everyone,
We are excited to invite you to a very special session of the Central NJ Data Science Meetup! For our upcoming event, we have a very focused presentation continuing our exploration of cutting-edge AI technologies.
Prayash Das will present ViTCap-R, a unified framework that bridges the gap between image captioning and cross-modal retrieval. Instead of treating these as independent problems, ViTCap-R integrates a Vision Transformer encoder, a patch-level attention LSTM decoder, and a dual-encoder contrastive retrieval module into a single pipeline. By attending over 196 spatial patch embeddings, the model produces highly interpretable captions grounded in specific image regions.
Prayash will walk us through the model's performance on standard benchmarks (Flickr8k and MS COCO) using metrics like BLEU, METEOR, and Recall@K. He will also share qualitative insights using attention heatmaps and t-SNE visualizations of the learned embedding space, followed by a discussion on the challenges of scaling and hard negative mining. He will make these advanced architecture concepts easy to follow, even for non-technical folks.
Event Details:

• Date: Saturday, July 18th, 2026.
• Location: East Brunswick Public Library (EBPL), 2 Jean Walling Civic Center, East Brunswick, NJ 08816.

### Our Featured Speaker:

• Prayash Das (In person): Demystifying Vision Transformers

### How to Join Us:

• In-Person (Preferred): We strongly encourage you to join us at the library for face-to-face networking and community building.
• Virtual Option: If you plan to attend remotely, you must register via the EBPL website to receive your unique Zoom link. (Select "Book as guest" if you are not an East Brunswick resident).

We look forward to an afternoon of shared learning and meaningful connections!