arXiv subjects

Sravan Chittupalli

Publications and source records attributed to Sravan Chittupalli.

3 recordsLinked to original sources

arXiv2026

Pi-HOC: Pairwise 3D Human-Object Contact Estimation

Resolving real-world human-object interactions in images is a many-to-many challenge, in which disentangling fine-grained concurrent physical contact is particularly difficult. Existing semantic contact estimation methods are either limited to single-human settings or require object geometries (e.g., meshes) in addition to the input image. Current state-of-the-art leverages powerful VLM for category-level semantics but struggles with multi-human scenarios and scales poorly in inference. We introduce Pi-HOC, a single-pass, instance-aware framework for dense 3D semantic contact prediction of all human-object pairs. Pi-HOC detects instances, creates dedicated human-object (HO) tokens for each pair, and refines them using an InteractionFormer. A SAM-based decoder then predicts dense contact on SMPL human meshes for each human-object pair. On the MMHOI and DAMON datasets, Pi-HOC significantly improves accuracy and localization over state-of-the-art methods while achieving 20x higher throughput. We further demonstrate that predicted contacts improve SAM-3D image-to-mesh reconstruction via a test-time optimization algorithm and enable referential contact prediction from language queries without additional training.

cs.CV↗

arXiv2025

BG-HOP: A Bimanual Generative Hand-Object Prior

In this work, we present BG-HOP, a generative prior that seeks to model bimanual hand-object interactions in 3D. We address the challenge of limited bimanual interaction data by extending existing single-hand generative priors, demonstrating preliminary results in capturing the joint distribution of hands and objects. Our experiments showcase the model's capability to generate bimanual interactions and synthesize grasps for given objects. We make code and models publicly available.

cs.CV↗

arXiv2022

DEEPFAKE CLI: Accelerated Deepfake Detection using FPGAs

Because of the availability of larger datasets and recent improvements in the generative model, more realistic Deepfake videos are being produced each day. People consume around one billion hours of video on social media platforms every day, and thats why it is very important to stop the spread of fake videos as they can be damaging, dangerous, and malicious. There has been a significant improvement in the field of deepfake classification, but deepfake detection and inference have remained a difficult task. To solve this problem in this paper, we propose a novel DEEPFAKE C-L-I (Classification-Localization-Inference) in which we have explored the idea of accelerating Quantized Deepfake Detection Models using FPGAs due to their ability of maximum parallelism and energy efficiency compared to generalized GPUs. In this paper, we have used light MesoNet with EFF-YNet structure and accelerated it on VCK5000 FPGA, powered by state-of-the-art VC1902 Versal Architecture which uses AI, DSP, and Adaptable Engines for acceleration. We have benchmarked our inference speed with other state-of-the-art inference nodes, got 316.8 FPS on VCK5000 while maintaining 93\% Accuracy.

cs.AR↗