MEDIA ANALYTICS
PROJECTS
PEOPLE
PUBLICATIONS
PATENTS
Media Analytics
Vijay Kumar B. G. is a Principal Researcher in the Media Analytics Department at NEC Laboratories America in San Jose, CA. He received his PhD in Computer and Information Sciences and Support Services from Queen Mary University London and his Master’s in Computer and Information Sciences and Support Services from the Indian Institute of Technology, Madras. Dr. Kumar’s work advances the integration of computer vision and language, with the goal of building intelligent systems that can interpret and interact with complex, real-world environments.
At NEC, he has contributed to research in agentic large language models, embodied AI, open-vocabulary perception, and foundational vision-language modeling. His recent projects include developing LLM-driven orchestration frameworks that convert natural-language task descriptions into executable workflows, as well as embodied AI systems that navigate and manipulate objects in unstructured spaces. He also works on open-vocabulary perception models that detect and describe previously unseen objects, and on creating general-purpose vision-language models that can be adapted to domains such as autonomous driving, law enforcement, and remote sensing. His research emphasizes efficiency, scalability, and reducing hallucination in AI outputs, ensuring these systems are practical for deployment.
He has published extensively in top-tier venues, including CVPR, ICCV, and NeurIPS, with recent contributions such as “DWIM: Towards Tool-aware Visual Reasoning” (2025). His work directly supports NEC’s mission to build perception and reasoning systems that improve safety, efficiency, and decision-making across industries.
Overview: We develop an agentic LLM to solve complex workflows by deploying a combination of computer vision, logic and compute modules. Based on a natural language task specification, our LLM generates a plan to accomplish the task using available tools.
Overview: We develop embodied agents for robotics applications that require exploration, navigation and transport in complex scenes. Our modular hierarchical transport policy builds a topological graph of the scene to perform exploration, then combined motion planning algorithms to reach point goals within explored locations with object navigation policies for moving towards semantic targets at unknown locations.
Overview: Our foundational models enable ubiquitous usage of computer vision across scenarios, applications and user preferences.
Overview: We develop open vocabulary perception methods that combine the power of vision and language to provide rich descriptions of objects in scenes, including their attributes, behaviors, relations and interactions.
Driving vision-language models (VLMs) must accurately understand scenes across diverse conditions defined by Operational Design Domains (ODDs), yet verificationremains sparse: many slices are missing, making empirical failure rates unreliable. We propose SLICESCORER, a deterministic scoring rule for
Visual reasoning (VR), which is crucial in many fields for enabling human-like visual understanding, remains highly challenging. Recently, compositional visual reasoning approaches, which leverage the reasoning abilities of large language models (LLMs) with integrated tools to solve problems, have shown
Spatio-temporal reasoning is essential in understanding real-world environments in various fields, eg, autonomous driving and sports analytics. Recent advances have improved the spatial reasoning ability of Vision-Language Models (VLMs) by introducing large-scale data, but these models still struggle
The development of Agentic LLMs for AI Orchestration represents a significant advancement in artificial intelligence. By seamlessly integrating computer vision, logic, and compute modules, our LLM is poised to revolutionize the way complex workflows are managed and executed. Supported by robust research
We propose a vision-LLM framework for automating development and deployment of computer vision solutions for pre-defined or custom-defined tasks. A foundational layer is proposed with a code-LLM AI orchestrator self-trained with reinforcement learning to create Python code based on its understanding
Recent studies have shown promising performance in open-vocabulary object detection (OVD) by utilizing pseudo labels (PLs) from pretrained vision and language models (VLMs). However, teacher-student self-training, a powerful and widely used paradigm to leverage PLs, is rarely explored for OVD.
Visual program synthesis is a promising approach to exploit the reasoning abilities of large language models for compositional computer vision tasks. Previous work has used few-shot prompting with frozen LLMs to synthesize visual programs. Training an LLM to write better visual programs is an attractive
The recent progress in language-based open-vocabulary object detection can be largely attributed to finding better ways of leveraging large-scale data with free-form text annotations. Training such models with a discriminative objective function has proven successful, but requires good positive and negative
The recent progress in language-based object detection with an open-vocabulary can be largely attributed to finding better ways of leveraging large-scale data with free-form text annotations. Training from image captions with grounded bounding boxes (ground truth or pseudo-labeled) enable the models
Visual question answering (VQA) has traditionally been treated as a single-step task where each question receives the same amount of effort, unlike natural human question-answering strategies. We explore a question decomposition strategy for VQA to overcome this limitation. We probe the ability of recently
Data augmentation techniques, such as image transformations and combinations, are highly effective at improving the generalization of computer vision models, especially when training data is limited. However, such techniques are fundamentally incompatible with differentially private learning approaches,
Although planning is a crucial component of the autonomous driving stack, researchers have yet to develop robust planning algorithms that are capable of safely handling the diverse range of possible driving scenarios. Learning-based planners suffer from overfitting and poor long-tail performance. On
NEC Labs America is proud to be a Silver Sponsor for NeurIPS 2023 in New Orleans from December 10-16. Visit our booth to meet our team and learn about our intern opportunities in machine learning, data science, media analytics and integrated systems. Also, our Vijay Kumar.B.G, Samuel Schulter & Manmohan
Language-based object detection is a promising direction towards building a natural interface to describe objects in images that goes far beyond plain category names. While recent methods show great progress in that direction, proper evaluation is lacking. With OmniLabel, we propose a novel task definition,
Recent studies show promising performance in open-vocabulary object detection (OVD) using pseudo labels (PLs) from pretrained vision and language models (VLMs). However, PLs generated by VLMs are extremely noisy due to the gap between the pretraining objective of VLMs and OVD, which blocks further advances
Finetuning a large vision language model (VLM) on a target dataset after large scale pretraining is a dominant paradigm in visual question answering (VQA). Datasets for specialized tasks such as knowledge-based VQA or VQA in non natural-image domains are orders of magnitude smaller than those for general-purpose
Building robust and generic object detection frameworks requires scaling to larger label spaces and bigger training datasets. However, it is prohibitively costly to acquire annotations for thousands of categories at a large scale. We propose a novel method that leverages the rich semantics available
Self-supervised vision-language pretraining from pure images and text with a contrastive loss is effective, but ignores fine-grained alignment due to a dual-stream architecture that aligns image and text representations only on a global level. Earlier, supervised, non-contrastive methods were capable