A
Arijit Ray
Researcher at Presidency University, Kolkata
Publications - 62
Citations - 918
Arijit Ray is an academic researcher from Presidency University, Kolkata. The author has contributed to research in topics: Mafic & Gabbro. The author has an hindex of 16, co-authored 57 publications receiving 731 citations. Previous affiliations of Arijit Ray include Steel Authority of India & Tata Institute of Fundamental Research.
Papers
More filters
Journal ArticleDOI
Grain-scale deformation in the Palaeoproterozoic Dongargarh Supergroup, central India: implications for shallow crustal deformation mechanisms from microstructural analysis
TL;DR: In this paper, the deformation was accomplished by a combination of pressure solution, microfracturing and dislocation creep processes in folded rocks of the Dongargarh Supergroup, central India.
Dissertation
The Art of Deep Connection - Towards Natural and Pragmatic Conversational Agent Interactions
TL;DR: A novel game called the Visual 20 Questions Game is introduced, where a machine tries to figure out a secret image a human has picked by having a natural language conversation with the human by using sequence-to-sequence learning and reinforcement learning, which demonstrate promise towards scalable and reasonable performances on both the tasks.
Journal ArticleDOI
Language-Guided Audio-Visual Source Separation via Trimodal Consistency
Reuben Tan,Arijit Ray,Andrea Burns,Bryan A. Plummer,Justin Salamon,Oriol Nieto,Bryan Russell,Ksenia Saenko +7 more
TL;DR: In this paper , a self-supervised approach for audio source separation in videos based on natural language queries is proposed, using only unlabeled video and audio pairs as training data.
Posted Content
Knowing What VQA Does Not: Pointing to Error-Inducing Regions to Improve Explanation Helpfulness.
TL;DR: The authors proposed Error Maps that clarify the error by highlighting image regions where the model is prone to err and thus, improve users' understanding of those cases, and further introduce a metric that simulates users' interpretation of explanations to evaluate their potential helpfulness to understand model correctness.
Journal ArticleDOI
COLA: How to adapt vision-language models to Compose Objects Localized with Attributes?
TL;DR: This paper proposed a text-to-image retrieval benchmark, Cola, to measure the lack of compositional capability of large vision-language models and explore modeling designs to adapt pre-trained language models to reason compositionally about multiple attributes attached to multiple objects.