Video Quality Assessment (VQA) strives to computationally emulate human perceptual judgments and has garnered significant attention given its widespread applicability. However, existing methodologies face two primary impediments: (1) limited proficiency in evaluating samples at quality extremes (e.g., severely degraded or near-perfect videos), and (2) insufficient sensitivity to nuanced quality variations arising from a misalignment with human perceptual mechanisms. Although vision-language models offer promising semantic understanding, their reliance on visual encoders pre-trained for high-level tasks often compromises their sensitivity to low-level distortions. To surmount these challenges, we propose the Restoration-Assisted Multi-modality VQA (RAM-VQA) framework. Uniquely, our approach leverages video restoration as a proxy to explicitly model distortion-sensitive features. The framework operates through two synergistic stages: a prompt learning stage that constructs a quality-aware textual space using triple-level references (degraded, restored, and pristine) derived from the restoration process, and a dual-branch evaluation stage that integrates semantic cues with technical quality indicators via spatio-temporal differential analysis. Extensive experiments demonstrate that RAM-VQA achieves state-of-the-art performance across diverse benchmarks, exhibiting superior capability in handling extreme-quality content while ensuring robust generalization.

Neural Radiance Fields (NeRF) show impressive performances in photo-realistic free-view rendering of scenes. Recent improvements such as TensoRF and ZipNeRF employ explicit models for faster optimization and rendering. However, all these radiance fields require a dense sampling of images in the given scene for effective training. Their performances degrade significantly when only a sparse set of views is available. Existing depth priors used to supervise the radiance fields are either sparse or suffer from generalization issues. We seek to learn scene-specific dense depth priors to regularize the radiance fields. Further, we desire a framework of regularizations that can work across different radiance field models. We observe that certain features of the radiance fields, such as positional encoding, number of decomposed tensor components or size of the hash table, cause overfitting in the sparse-input scenario. We design augmented models by reducing the capacity of these features and train them along with the main radiance field. These augmented models learn simpler solutions, which estimate better depth in certain regions. By supervising the main radiance field with such depths, we significantly improve the performance of the radiance fields on popular forward-facing and 360 degrees datasets by employing the above regularization.

Despite modeling temporal motion, dynamic 3D Gaussian Splatting (3DGS) methods still inherit a static densification strategy that is ill-suited for dynamic scenes. This neglect of temporal behavior leads to under-reconstructed and blurry dynamic regions, as short-lived Gaussians receive sparse supervision and fail to densify effectively. We propose a Visibility-Aware Densification (VAD) framework that integrates temporal visibility into the densification process, ensuring that Gaussians are refined based on their actual temporal presence. A Temporally-Adaptive Thresholding (TAT) mechanism further adjusts each Gaussian's densification threshold according to its temporal lifespan, promoting balanced refinement of both static and dynamic regions. Finally, a Temporal Offset Warping (TOW) design enhances deformation capacity around temporal centers, extending the lifespan of highly dynamic Gaussians and facilitating more effective densification. Our approach achieves substantial improvements in the visual quality of dynamic regions, outperforming existing methods across three dynamic multi-view benchmark datasets. Moreover, the proposed VAD module generalizes across diverse dynamic 3DGS methods, consistently improving dynamic reconstruction as a plug-and-play component.

Existing text-to-video (T2V) evaluation benchmarks, such as VBench and EvalCrafter, suffer from two limitations. (i) While the emphasis is on subject-centric prompts or static camera scenes, camera motion essential for producing cinematic shots and existing metrics under dynamic motion are largely unexplored. (ii) These benchmarks typically aggregate video-level scores into a single model-level score for ranking generative models. Such aggregation, however, overlook video-level evaluation, which is vital to selecting the better video among the candidate videos generated for a given prompt. To address these gaps, we introduce DynamicEval, a benchmark consisting of systematically curated prompts emphasizing dynamic camera motion, paired with 45k human annotations on video pairs from 3k videos generated by ten T2V models. DynamicEval evaluates two key dimensions of video quality: background scene consistency and foreground object consistency. For background scene consistency, we obtain the interpretable error maps based on the Vbench motion smoothness metric. We observe that while the Vbench motion smoothness metric shows promising alignment with human judgments, it fails in two cases: occlusions/disocclusions arising from camera and foreground object movements. Building on this, we propose a new background consistency metric that leverages object error maps to correct two failure cases in a principled manner. Our second innovation is the introduction of a foreground consistency metric that tracks points and their neighbors within each object instance to assess object fidelity. Extensive experiments demonstrate that our proposed metrics achieve stronger correlations with human preferences at both the video level and the model level (an improvement of more than 2

Perceptual assessment of user-generated content videos is an important problem that impacts viewing experience of millions of users. Current no-reference video quality assessment (NR-VQA) algorithms require a large amount of human annotated videos. In this work, we address this problem by specifically designing a dual-model based Semi-supervised Learning (SSL) method for NR-VQA. The first model is based on a popular vision-language model namely CLIP, where we adapt the visual encoder to capture high-level semantic quality information through quality-relevant text prompts. A second model learns complementary low-level spatio-temporal quality using a 3D vision transformer and video fragments. We enable intelligent knowledge transfer between the high-level vision-language and low-level vision-transformer model to pseudo-label the unlabelled videos. Our unified model outperforms existing state-of-the-art SSL methods for VQA across popular VQA databases including inter-database settings.

Novel view synthesis involves generating unseen perspectives of a scene based on videos captured from limited viewpoints. Learning scene representations of such videos containing dynamic scene elements introduces several challenges in modeling the motion. Existing models typically require dense viewpoint coverage to produce high-quality renderings, and their performance degrades significantly when the number of viewpoints is reduced. In this work, we explore the challenges associated with volumetric motion modeling for synthesizing novel views of dynamic scenes from only a few fixed viewpoints. Specifically, we address the limitations of unidirectional motion models, which often result in a many-to-one mapping of scene points. We enforce cyclic flow consistency with the help of bidirectional motion fields to achieve superior reconstruction of novel views of dynamic scenes. Moreover, the bi-directional motion field design allows us to track object motion in the synthesized views.

With the rise of marine exploration, underwater imaging has gained significant attention as a research topic. Underwater video enhancement has become crucial for real-time computer vision tasks in marine exploration. However, most existing methods focus on enhancing individual frames and neglect video temporal dynamics, leading to visually poor enhancements. Furthermore, the lack of ground-truth references limits the use of abundant available underwater video data in many applications. To address these issues, we propose a two-stage framework for enhancing underwater videos. The first stage uses a denoising diffusion probabilistic model to learn a generative prior from unlabeled data, capturing robust and descriptive feature representations. In the second stage, this prior is incorporated into a physics-based image formulation for spatial enhancement, while also enforcing temporal consistency between video frames. Our method enables real-time and computationally-efficient processing of high-resolution underwater videos at lower resolutions, and offers efficient enhancement in the presence of diverse water-types. Extensive experiments on four datasets show that our approach generalizes well and outperforms existing enhancement methods. Our code is available at github.com/suhas-srinath/undive.

Multispectral remote sensing images suffer from a wide range of degradations due to noise, blur, haze, compression, striping, and other distortions. While there is rich literature in the restoration of such multispectral images, the quality assessment (QA) of multispectral images has received much less attention. In addition, often a reference image is not available, which motivates the study of no-reference (NR) image QA (IQA). While several algorithms for NR-IQA have been designed for natural images, their relevance for multispectral images has not been validated. The major impediment to achieving this is the lack of a subjective QA database for validation. We make three contributions in our work. We first design a multispectral IQA database, namely, the multispectral remote sensing image database (MRSID), with images suffering from various distortions and conduct a subjective study to obtain pairwise image quality rankings. We then benchmark several NR-IQA methods developed for natural images on MRSID. We show that recent vision-language-based models can yield excellent performance when fine-tuned on our dataset, especially when multispectral images are trained with images pooled from all bands, achieving an improvement of 15.8% in pairwise preference accuracy over single-band training. Finally, we investigate the reasons for such superior performance and show that improving the alignment of features across different spectral bands can further improve performance, yielding an additional 3.1% gain in pairwise preference accuracy on the cross-scene evaluation split, indicating improved generalization.

3D Gaussian Splatting (3DGS) has shown impressive results in real-time novel view synthesis. However, it often struggles under sparse-view settings, producing undesirable artifacts such as floaters, inaccurate geometry, and overfitting due to limited observations. We find that a key contributing factor is uncontrolled densification, where adding Gaussian primitives rapidly without guidance can harm geometry and cause artifacts. We propose AD-GS, a novel alternating densification framework that interleaves high and low densification phases. During high densification, the model densifies aggressively, followed by photometric loss based training to capture fine-grained scene details. Low densification then primarily involves aggressive opacity pruning of Gaussians followed by regularizing their geometry through pseudo-view consistency and edge-aware depth smoothness. This alternating approach helps reduce overfitting by carefully controlling model capacity growth while progressively refining the scene representation. Extensive experiments on challenging datasets demonstrate that AD-GS significantly improves rendering quality and geometric consistency compared to existing methods.

Designing learning-based no-reference (NR) video quality assessment (VQA) algorithms for camera-captured videos is cumbersome due to the requirement of a large number of human annotations of quality. In this work, we propose a semi-supervised learning (SSL) framework exploiting many unlabelled and very limited amounts of labelled authentically distorted videos. Our main contributions are two-fold. Leveraging the benefits of consistency regularization and pseudo-labelling, our SSL model generates pairwise pseudo-ranks for the unlabelled videos using a student-teacher model on strongweak augmented videos. We design the strong-weak augmentations to be quality invariant to use the unlabelled videos effectively in SSL. The generated pseudo-ranks are used along with the limited labels to train our SSL model. Our primary focus in SSL for NR VQA is to learn the mapping from video feature representations to the quality scores. We compare various feature extraction methods and show that our SSL framework can lead to improved performance on these features. In addition to the existing features, we present a spatial and temporal feature extraction method based on predicting spatial and temporal entropic differences. We show that these features help achieve a robust performance when trained with limited data providing a better baseline to apply SSL. Extensive experiments on three popular VQA datasets demonstrate that a combination of our novel SSL approach and features achieves an impressive performance in terms of correlation with human perception, even though the number of human-annotated videos may be limited.

The design of image and video quality assessment (QA) algorithms is extremely important to benchmark and calibrate user experience in modern visual systems. A major drawback of the state-of-the-art QA methods is their limited ability to generalize across diverse image and video datasets with reasonable distribution shifts. In this work, we leverage the denoising process of diffusion models for generalized image QA (IQA) and video QA (VQA) by understanding the degree of alignment between learnable quality-aware text prompts and images or video frames. In particular, we learn cross-attention maps from intermediate layers of the denoiser of latent diffusion models (LDMs) to capture quality-aware representations of images or video frames. Since applying text-to-image LDMs for every video frame is computationally expensive for videos, we only estimate the quality of a frame-rate sub-sampled version of the original video. To compensate for the loss in motion information due to frame-rate sub-sampling, we propose a novel temporal quality modulator. Our extensive cross-database experiments across various user-generated, synthetic, low-light, frame-rate variation, ultra high definition, and streaming content-based databases show that our model can achieve superior generalization in both IQA and VQA.

Perceptual quality assessment of user generated content (UGC) videos is challenging due to the requirement of large scale human annotated videos for training. In this work, we address this challenge by first designing a self-supervised Spatio-Temporal Visual Quality Representation Learning (ST-VQRL) framework to generate robust quality aware features for videos. Then, we propose a dual-model based Semi Supervised Learning (SSL) method specifically designed for the Video Quality Assessment (SSL-VQA) task, through a novel knowledge transfer of quality predictions between the two models. Our SSL-VQA method uses the ST-VQRL backbone to produce robust performances across various VQA datasets including cross-database settings, despite being learned with limited human annotated videos. Our model improves the state-of-the-art performance when trained only with limited data by around 10%, and by around 15% when unlabelled data is also used in SSL. Source codes and checkpoints are available at https://github.com/Shankhanil006/SSL-VQA.

Designing a 3D representation of a dynamic scene for fast optimization and rendering is a challenging task. While recent explicit representations enable fast learning and rendering of dynamic radiance fields, they require a dense set of input viewpoints. In this work, we focus on learning a fast representation for dynamic radiance fields with sparse input viewpoints. However, the optimization with sparse input is under-constrained and necessitates the use of motion priors to constrain the learning. Existing fast dynamic scene models do not explicitly model the motion, making them difficult to be constrained with motion priors. We design an explicit motion model as a factorized 4D representation that is fast and can exploit the spatio-temporal correlation of the motion field. We then introduce reliable flow priors including a combination of sparse flow priors across cameras and dense flow priors within cameras to regularize our motion model. Our model is fast, compact and achieves very good performance on popular multi-view dynamic scene datasets with sparse input viewpoints. The source code for our model can be found on our project page: https://nagabhushansn95.github.io/publications/2024/RF-DeRF.html.

One of the major challenges in no-reference (NR) image quality assessment (IQA) is the ability to generalize to diverse quality assessment applications. Recently, multi-modal vision-language models are found to be very promising in this direction. They are beginning to form a part of several state of the art NR IQA methods. On the other hand, multi-modal large language models (LLMs) are increasingly being studied for various computer vision applications including IQA. In this work, we perform a thorough study of the ability of multi-modal LLMs for NR IQA by training some of its components and testing for its generalizability. In particular, we keep the LLM frozen and learn parameters corresponding to the querying transformer, LLM prompt and some layers that process the embedding output by the LLM. We observe that some of these components offer a generalization performance far superior to any existing NR IQA algorithm.

No-reference (NR) image quality assessment (IQA) is an important tool in enhancing the user experience in diverse visual applications. A major drawback of state-of-the-art NR-IQA techniques is their reliance on a large number of human annotations to train models for a target IQA application. To mitigate this requirement, there is a need for unsupervised learning of generalizable quality representations that capture diverse distortions. We enable the learning of low-level quality features agnostic to distortion types by introducing a novel quality-aware contrastive loss. Further, we leverage the generalizability of vision-language models by fine-tuning one such model to extract high-level image quality information through relevant text prompts. The two sets of features are combined to effectively predict quality by training a simple regressor with very few samples on a target dataset. Additionally, we design zero-shot quality predictions from both pathways in a completely blind setting. Our experiments on diverse datasets encompassing various distortions show the generalizability of the features and their superior performance in the data-efficient and zero-shot settings.

Neural radiance fields (NeRF) have achieved impressive performances in view synthesis by encoding neural representations of a scene. However, NeRFs require hundreds of images per scene to synthesize photo-realistic novel views. Training them on sparse input views leads to overfitting and incorrect scene depth estimation resulting in artifacts in the rendered novel views. Sparse input NeRFs were recently regularized by providing dense depth estimated from pre-trained networks as supervision, to achieve improved performance over sparse depth constraints. However, we find that such depth priors may be inaccurate due to generalization issues. Instead, we hypothesize that the visibility of pixels in different input views can be more reliably estimated to provide dense supervision. In this regard, we compute a visibility prior through the use of plane sweep volumes, which does not require any pre-training. By regularizing the NeRF training with the visibility prior, we successfully train the NeRF with few input views. We reformulate the NeRF to also directly output the visibility of a 3D point from a given viewpoint to reduce the training time with the visibility constraint. On multiple datasets, our model outperforms the competing sparse input NeRF models including those that use learned priors. The source code for our model can be found on our project page: https://nagabhushansn95.github.io/publications/2023/ViP-NeRF.html.

Convolutional neural networks have been successful in restoring images captured under poor illumination conditions. Nevertheless, such approaches require a large number of paired low-light and ground truth images for training. Thus, we study the problem of semi-supervised learning for low-light image restoration when limited low-light images have ground truth labels. Our main contributions in this work are twofold. We first deploy an ensemble of low-light restoration networks to restore the unlabeled images and generate a set of potential pseudo-labels. We model the contrast distortions in the labeled set to generate different sets of training data and create the ensemble of networks. We then design a contrastive selfsupervised learning based image quality measure to obtain the pseudo-label among the images restored by the ensemble. We show that training the restoration network with the pseudo-labels allows us to achieve excellent restoration performance even with very few labeled pairs. We conduct extensive experiments on three popular low-light image restoration datasets to show the superior performance of our semi-supervised low-light image restoration compared to other approaches. Project page is available at https://github.com/sameerIISc/SSL-LLR.

The quality assessment (QA) of camera captured authentically distorted images is important on account of its ubiquitous applications and challenging due to the lack of a reference. While there exists a plethora of supervised no reference (NR) image QA (IQA) algorithms, there is a need to study unsupervised or opinion unaware algorithms on account of their superior generalization performance. We explore self-supervised learning (SSL) for the feature design on authentically distorted images to predict quality without training on human labels. While SSL on synthetic distortions has recently shown promise, there is a need to enrich the feature learning on authentic distortions. The key challenge in achieving this is in the learning of quality sensitive features with mitigated content dependence. We design a self-supervised contrastive learning approach which only requires positives and introduce a content separation loss by estimating a bound on the mutual information between the features learnt and the content information. We show on multiple authentically distorted datasets that our self-supervised features can predict image quality by comparing with a corpus of pristine images and achieve state-of-the-art performance. §

Neural Radiance Fields (NeRF) show impressive performance for the photorealistic free-view rendering of scenes. However, NeRFs require dense sampling of images in the given scene, and their performance degrades significantly when only a sparse set of views are available. Researchers have found that supervising the depth estimated by the NeRF helps train it effectively with fewer views. The depth supervision is obtained either using classical approaches or neural networks pre-trained on a large dataset. While the former may provide only sparse supervision, the latter may suffer from generalization issues. As opposed to the earlier approaches, we seek to learn the depth supervision by designing augmented models and training them along with the NeRF. We design augmented models that encourage simpler solutions by exploring the role of positional encoding and view-dependent radiance in training the few-shot NeRF. The depth estimated by these simpler models is used to supervise the NeRF depth estimates. Since the augmented models can be inaccurate in certain regions, we design a mechanism to choose only reliable depth estimates for supervision. Finally, we add a consistency loss between the coarse and fine multi-layer perceptrons of the NeRF to ensure better utilization of hierarchical sampling. We achieve state-of-the-art view-synthesis performance on two popular datasets by employing the above regularizations. The source code for our model can be found on our project page: https://nagabhushansn95.github.io/publications/2023/SimpleNeRF.html

While the design of blind image quality assessment (IQA) algorithms has improved significantly, the distribution shift between the training and testing scenarios often leads to a poor performance of these methods at inference time. This motivates the study of test time adaptation (TTA) techniques to improve their performance at inference time. Existing auxiliary tasks and loss functions used for TTA may not be relevant for quality-aware adaptation of the pre-trained model. In this work, we introduce two novel quality-relevant auxiliary tasks at the batch and sample levels to enable TTA for blind IQA. In particular, we introduce a group contrastive loss at the batch level and a relative rank loss at the sample level to make the model quality aware and adapt to the target data. Our experiments reveal that even using a small batch of images from the test distribution helps achieve significant improvement in performance by updating the batch normalization statistics of the source model.