abstract / bibtex / proceedings / arXiv / Code

We aim to tackle sparse-view reconstruction of a 360° 3D scene using priors from latent diffusion models (LDM). The sparse-view setting is ill-posed and underconstrained, especially for scenes where the camera rotates 360 degrees around a point, as no visual information is available beyond some frontal views focused on the central object(s) of interest. In this work, we show that pretrained 2D diffusion models can strongly improve the reconstruction of a scene with low-cost fine-tuning. Specifically, we present SparseSplat360 (Sp2360), a method that employs a cascade of in-painting and artifact removal models to fill in missing details and clean novel views. Due to superior training and rendering speeds, we use an explicit scene representation in the form of 3D Gaussians over NeRF-based implicit representations. We propose an iterative update strategy to fuse generated pseudo novel views with existing 3D Gaussians fitted to the initial sparse inputs. As a result, we obtain a multi-view consistent scene representation with details coherent with the observed inputs. Our evaluation on the challenging Mip-NeRF360 dataset shows that our proposed 2D to 3D distillation algorithm considerably improves the performance of a regularized version of 3DGS adapted to a sparse-view setting and outperforms existing sparse-view reconstruction methods in 360° scene reconstruction. Qualitatively, our method generates entire 360° scenes from as few as 9 input views, with a high degree of foreground and background detail.

        @inproceedings{
          paul2024sp,
          title={Sp2360: Sparse-view 360{\textopenbullet} Scene Reconstruction using Cascaded 2D Diffusion Priors},
          author={Soumava Paul and Christopher Wewer and Bernt Schiele and Jan Eric Lenssen},
          booktitle={ECCV 2024 Workshop on Wild 3D: 3D Modeling, Reconstruction, and Generation in the Wild},
          year={2024},
          url={https://openreview.net/forum?id=XuNhNyHHwK}
        }
      

abstract / bibtex / arXiv / Code

Image retrieval under generalized test scenarios has gained significant momentum in literature, and the recently proposed protocol of Universal Cross-domain Retrieval is a pioneer in this direction. A common practice in any such generalized classification or retrieval algorithm is to exploit samples from many domains during training to learn a domain-invariant representation of data. Such criterion is often restrictive, and thus in this work, for the first time, we explore the generalized retrieval problem in a data-efficient manner. Specifically, we aim to generalize any pre-trained cross-domain retrieval network towards any unknown query domain/category, by means of adapting the model on the test data leveraging self-supervised learning techniques. Toward that goal, we explored different self-supervised loss functions (for example, RotNet, JigSaw, Barlow Twins, etc.) and analyze their effectiveness for the same. Extensive experiments demonstrate the proposed approach is simple, easy to implement, and effective in handling data-efficient UCDR.

      @misc{paul2023testtime,
        title={Test-time Training for Data-efficient UCDR}, 
        author={Soumava Paul and Titir Dutta and Aheli Saha and Abhishek Samanta and Soma Biswas},
        year={2023},
        eprint={2208.09198},
        archivePrefix={arXiv},
        primaryClass={cs.CV}
      }
      

abstract / bibtex / proceedings / arXiv / Code / Video / slides / poster

In this work, for the first time, we address the problem of universal cross-domain retrieval, where the test data can belong to classes or domains which are unseen during training. Due to dynamically increasing number of categories and practical constraint of training on every possible domain, which requires large amounts of data, generalizing to both unseen classes and domains is important. Towards that goal, we propose SnMpNet (Semantic Neighbourhood and Mixture Prediction Network), which incorporates two novel losses to account for the unseen classes and domains encountered during testing. Specifically, we introduce a novel Semantic Neighborhood loss to bridge the knowledge gap between seen and unseen classes and ensure that the latent space embedding of the unseen classes is semantically meaningful with respect to its neighboring classes. We also introduce a mix-up based supervision at image-level as well as semantic-level of the data for training with the Mixture Prediction loss, which helps in efficient retrieval when the query belongs to an unseen domain. These losses are incorporated on the SE-ResNet50 backbone to obtain SnMpNet. Extensive experiments on two large-scale datasets, Sketchy Extended and DomainNet, and thorough comparisons with state-of-the-art justify the effectiveness of the proposed model.

      @InProceedings{Paul_2021_ICCV,
        author    = {Paul, Soumava and Dutta, Titir and Biswas, Soma},
        title     = {Universal Cross-Domain Retrieval: Generalizing Across Classes and Domains},
        booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
        month     = {October},
        year      = {2021},
        pages     = {12056-12064}
      }
      

abstract / bibtex / proceedings / arXiv / Code / Video / slides

Singing Voice Detection (SVD) has been an active area of research in music information retrieval (MIR). Currently, two deep neural network-based methods, one based on CNN and the other on RNN, exist in literature that learn optimized features for the voice detection (VD) task and achieve state-of-the-art performance on common datasets. Both these models have a huge number of parameters (1.4M for CNN and 65.7K for RNN) and hence not suitable for deployment on devices like smartphones or embedded sensors with limited capacity in terms of memory and computation power. The most popular method to address this issue is known as knowledge distillation in deep learning literature (in addition to model compression) where a large pre-trained network known as the teacher is used to train a smaller student network. Given the wide applications of SVD in music information retrieval, to the best of our knowledge, model compression for practical deployment has not yet been explored. In this paper, efforts have been made to investigate this issue using both conventional as well as ensemble knowledge distillation techniques.

      @inproceedings{paul21b_interspeech,
        author={Soumava Paul and Gurunath Reddy M and K. Sreenivasa Rao and Partha Pratim Das},
        title={{Knowledge Distillation for Singing Voice Detection}},
        year=2021,
        booktitle={Proc. Interspeech 2021},
        pages={4159--4163},
        doi={10.21437/Interspeech.2021-636}
      }
      

abstract / bibtex / proceedings / arXiv / Code / Video / slides

Zero-shot learning (ZSL) algorithms typically work by exploiting attribute correlations to make predictions for unseen classes. However, these correlations do not remain intact at test time in most practical settings, and the resulting change in these correlations leads to adverse effects on zero-shot learning performance. In this paper, we present a new paradigm for ZSL that: (i) utilizes the class-attribute mapping of unseen classes to estimate the change in target distribution (target shift), and (ii) propose a novel technique called grouped Adversarial Learning (gAL) to reduce negative effects of this shift. Our approach is widely applicable for several existing ZSL algorithms, including those with implicit attribute predictions. We apply the proposed technique (gAL) on three popular ZSL algorithms: ALE, SJE, and DEVISE, and show performance improvements on 4 popular ZSL datasets: AwA2, aPY, CUB, and SUN.

      @InProceedings{Chemmengath_2021_ICCV,
        author    = {Chemmengath, Saneem A. and Paul, Soumava and Bharadwaj, Samarth and Samanta, Suranjana and Sankaranarayanan, Karthik},
        title     = {Addressing Target Shift in Zero-Shot Learning Using Grouped Adversarial Learning},
        booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
        month     = {October},
        year      = {2021},
        pages     = {2368-2377}
      }
      

abstract / bibtex / proceedings / Code / slides

Deep learning models trained in natural images are commonly used for different classification tasks in the medical domain. Generally, very high dimensional medical images are down-sampled by using interpolation techniques before feeding them to deep learning models that are ImageNet compliant and accept only low-resolution images of size 224 x 224 px. This popular technique may lead to the loss of key information thus hampering the classification. Significant pathological features in medical images typically being small sized and highly affected. To combat this problem, we introduce a convolutional neural network (CNN) based classification approach which learns to reduce the resolution of the image using an autoencoder and at the same time classify it using another network, while both the tasks are trained jointly. This algorithm guides the model to learn essential representations from high-resolution images for classification along with reconstruction. We have used the publicly available dataset of chest x-rays to evaluate this approach and have outperformed state-of-the-art on test data. Besides, we have experimented with the effects of different augmentation approaches in this dataset and report baselines using some well known ImageNet class of CNNs.

      @inproceedings{ranjan2018jointly,
        title={Jointly learning convolutional representations to compress radiological images and classify thoracic diseases in the compressed domain},
        author={Ranjan, Ekagra and Paul, Soumava and Kapoor, Siddharth and Kar, Aupendu and Sethuraman, Ramanathan and Sheet, Debdoot},
        booktitle={Proceedings of the 11th Indian Conference on computer vision, graphics and image processing},
        pages={1--8},
        year={2018}
      }
      

project page / Code

Our submission for the Rendering Competiton of the Computer Graphics-I course at UdS (Winter Term 2022/23). Finished in the A-tier with 80% points, narrowly missing out on the podium places.

Report / Code

Our submission for the Inter-Hall Hardware Modelling Competiton at IIT Kgp in 2019. We built a robotic exoskeleton system for assisting patients with muscular atrophy. Secured 4^th place out of 15 teams.