Hritam Basak.

We present an end-to-end video generation and policy-learning pipeline for exocentric-to-egocentric viewpoint translation, enabling transformative progress in imitation learning and advancing the frontier of embodied AI.

@inproceedings{basak2026exo2ego,
  title = {Exo2EgoPolicy: Pose-Aligned Cross-View Policy Learning},
  author = {Basak, Hritam and Tabatabaee, Hadi and Gayaka, Shreekant and Yang, Xin and Kuo, Cheng-Hao and Yin, Zhaozheng and Sun, Min},
  booktitle = {ECCV},
  year = {2026}
}

We propose a split-latent hierarchical diffusion framework that disentangles deterministic geometric priors (fixed latent) from stochastic scene-level context (diffused latent) for reliable affordance-grasp prediction in cluttered, partially observed environments. A Global Conditional Diffusion produces the stochastic latent from RGB context, a Local Conditional Diffusion predicts dense point-wise affordance and grasp fields, and a Harmonic Grasp Field enforces smooth, obstacle-aware grasp manifolds, yielding physically consistent and uncertainty-aware grasp generation.

@inproceedings{basak2026fixed,
  title = {Fixed Reality, Diffused Possibility: Disentangling Stochastic and Deterministic Latent for Cluttered Grasping},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {ECCV},
  year = {2026}
}

We introduce PhysioSplat, a physics-informed dynamic Gaussian Splatting framework for surgical scene reconstruction, embedding physical priors into deformable tissue modeling for geometrically coherent, real-time endoscopic rendering.

@inproceedings{basak2026physiosplat,
  title = {PhysioSplat: Physics-Informed Dynamic Gaussian Splatting for Surgical Scene Reconstruction},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {MICCAI},
  year = {2026}
}

We devise a Visual Iterative Self-Prompting algorithm with Neural Affordance Field supervision for language-guided 3D affordance estimation, enabling robots to ground language instructions in 3D actionable regions.

@inproceedings{basak2025vispla,
  title = {ViSPLA: Visual Iterative Self-Prompting for Language-Guided 3D Affordance Learning},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {NeurIPS},
  year = {2025}
}

3D object generation from a single unposed RGB image is essential for robotic perception... Our approach achieves state-of-the-art 3D reconstruction quality, significantly improving robotic perception pipelines.

@inproceedings{basak2025enhancing,
  title = {Enhancing Single Image to 3D Generation using Gaussian Splatting and Hybrid Diffusion Priors},
  author = {Basak, Hritam and Tabatabaee, Hadi and Gayaka, Shreekant and Li, Ming-Feng and Yang, Xin and Kuo, Cheng-Hao and Sen, Arnie and Sun, Min and Yin, Zhaozheng},
  booktitle = {IROS},
  year = {2025}
}

Deformable tissue reconstruction in endoscopy is vital for surgery... D4Recon achieves superior geometric coherence and photorealism with real-time rendering speed, outperforming existing methods.

@inproceedings{basak2025d4recon,
  title = {D4Recon: Dual-stage Deformation and Dual-scale Depth Guidance for Endoscopic Reconstruction},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {MICCAI},
  year = {2025}
}

We propose UA-Pose, an uncertainty-aware approach for 6D object pose estimation... Experimental results demonstrate significant performance improvements.

@inproceedings{li2025ua,
  title = {UA-Pose: Uncertainty-aware 6D Object Pose Estimation and Online Object Completion with Partial References},
  author = {Li, Ming-Feng and Yang, Xin and Wang, Fu-En and Basak, Hritam and Sun, Yuyin and Gayaka, Shreekant and Sun, Min and Kuo, Cheng-Hao},
  booktitle = {CVPR},
  year = {2025}
}

We propose the first language-guided SSDA setting for semantic segmentation. We harness the semantic generalization capabilities inherent in vision-language models (VLMs).

@inproceedings{basak2025semidavil,
  title = {SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {CVPR},
  year = {2025}
}

In pursuit of greater generalizability and robustness, we present an SSDA framework with a new episodic learning strategy: "learn, forget, then learn more".

@inproceedings{basak2024forget,
  title = {Forget More to Learn More: Domain-specific Feature Unlearning for Semi-supervised and Unsupervised Domain Adaptation},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {ECCV},
  year = {2024}
}

We propose a generative latent search paradigm to reconstruct the closest clone of every target image from the source latent space... We empirically validate the efficacy of our proposed method.

@inproceedings{basak2024quest,
  title = {Quest for Clone: Test-Time Domain Adaptation for Medical Image Segmentation by Searching the Closest Clone in Latent Space},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {MICCAI},
  year = {2024}
}

We propose a novel semi-supervised patch-based CL framework for medical image segmentation... Experimental analysis reveals the superiority of our proposed method.

@inproceedings{basak2023pseudo,
  title = {Pseudo Label Guided Contrastive Learning for Semi-supervised Medical Image Segmentation},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {CVPR},
  year = {2023}
}

We investigate semi-supervised domain adaptation (SSDA) for medical image segmentation... Our proposed method outperforms the SoTA methods.

@inproceedings{basak2023semi,
  title = {Semi-supervised Domain Adaptive Medical Image Segmentation through Consistency Regularized Disentangled Contrastive Learning},
  author = {Basak, Hritam and Yin, Zhaozheng},
  booktitle = {MICCAI},
  year = {2023}
}

To alleviate the shortcoming of class imbalance, we propose maintaining a confidence array that records class-wise performance during training.

@inproceedings{basak2022addressing,
  title = {Addressing Class Imbalance in Semi-supervised Image Segmentation: A Study on Cardiac MRI},
  author = {Basak, Hritam and Ghosal, Sagnik and Sarkar, Ram},
  booktitle = {MICCAI},
  year = {2022}
}

We introduce a novel regularization strategy involving interpolation-based mixing for semi-supervised medical image segmentation.

@inproceedings{basak2022exceedingly,
  title = {An Exceedingly Simple Consistency Regularization Method for Semi-Supervised Medical Image Segmentation},
  author = {Basak, Hritam and Bhattacharya, Rajarshi and Hussain, Rukhshanda and Chatterjee, Agniv},
  booktitle = {ISBI},
  year = {2022}
}