Science AI Journal
Computer Science · Research topic

Open research questions in Advanced Vision and Imaging

122 unresolved questions extracted from the limitations and future-work sections of 533 Advanced Vision and Imaging papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • It can be observed that although reliable 2D instance segmentation masks have been provided by off-the-shelf 2D models, the network may fail in difficult situations, which indicates that our method is mainly limited by the sparse and noisy nature of radar data itself, rather than insufficient support from 2D semantics.

    Weakly Supervised Cross-Modal Learning for 4D Radar Scene Flow Estimation · 2026

  • However, the principles governing how these neurons filter their inputs to generate appropriate responses remain unclear.

    Dendritic architecture enables de novo computation of salient motion in the superior colliculus · 2025 · DOI

  • The gap in current research is the requirement of a large number of input views, which can be resource-intensive. The gap is the inability to render photorealistic street-view panoramas with a small number of input views.

    Seeing Through Satellite Images at Street Views · 2026 · DOI

  • Exploring the use of Lite3R in various applications, such as robotics and computer vision. Investigating the potential of Lite3R to improve the efficiency of 3D reconstruction models in real-world scenarios.

    Lite3R: A Model-Agnostic Framework for Efficient Feed-Forward 3D Reconstruction · 2026

  • Modern 3D transformer pipelines face significant challenges, including dense multi-view attention and low-precision execution. Improving the efficiency of 3D reconstruction models is crucial for practical deployment.

    Lite3R: A Model-Agnostic Framework for Efficient Feed-Forward 3D Reconstruction · 2026

  • The paper identifies the need for more diverse modalities and more accurate reconstruction. The authors discuss the challenge of free-viewpoint synthesis and long-context generation. The survey highlights the potential misuse of generative capabilities of feed-forward 3D reconstruction models.

    Advances in Feed‐Forward 3D Reconstruction and View Synthesis: A Survey · 2026 · DOI

  • Improving model generalizability to reduce computational costs - Developing advanced detection models to distinguish between generated and real content - Establishing new regulations as 3D reconstruction technologies become increasingly accessible

    Advances in Feed‐Forward 3D Reconstruction and View Synthesis: A Survey · 2026 · DOI

  • Further improvement of the method for scenes with very large camera movements or tissue deformations. Exploration of real-time applications of the method. Investigation of potential applications in medical imaging and computer vision.

    Endo-PairGS: pair priors for dynamic endoscopic scene reconstruction · 2026 · DOI

  • Existing methods struggle with challenging scenes, such as those with deformable tissues and surgical tool occlusion. Inferior initialization of 3D Gaussians in dynamic endoscopic scenes. Limited computational efficiency during training and inference.

    Endo-PairGS: pair priors for dynamic endoscopic scene reconstruction · 2026 · DOI

  • Conventional RGB-based stereo matching algorithms suffer from ambiguities caused by metamerism and weak textures. The lack of a co-aperture system and a Measure-and-Complete strategy for generating dense pseudo-ground truth.

    4D-Aware Stereo Matching via Implicit Spectral Reconstruction with Multi-modal Training and RGB-only Deployment · 2026 · DOI

  • First, the spectral reconstruction is restricted to the 450–700 nm visible band; its performance in the near-infrared and short-wave infrared regimes—where many distinctive material fingerprints reside—remains to be validated. 8 m), and generalization to medium- and long-range scenarios remains an open question.

    4D-Aware Stereo Matching via Implicit Spectral Reconstruction with Multi-modal Training and RGB-only Deployment · 2026 · DOI

  • Existing generation paradigms struggle to satisfy the requirements of whole-house tours simultaneously. There is a gap in generating consistent whole-house panoramas with high-fidelity and material consistency.

    PanoWorld: A Generative Spatial World Model for Consistent Whole-House Panorama Synthesis · 2026

  • Exploring metrics that quantify the similarity gap between correct and uncertain matches. Evaluating the performance of the proposed approach in different environments and conditions.

    Patch Ensembles for Robust Salmon Re-Identification with Weak Trajectory Labels · 2026

  • The lack of labeled data for large-scale salmon re-identification. The high accuracy requirements imposed by commercial net-pens. The need for a realistic evaluation setup for salmon re-identification.

    Patch Ensembles for Robust Salmon Re-Identification with Weak Trajectory Labels · 2026

  • Existing methods typically trade off generalization, geometric fidelity, and efficiency. There is a need for an automated correspondence framework that is accurate, robust, and computationally efficient.

    SGSoft: Learning Fused Semantic-Geometric Features for 3D Shape Correspondence via Template-Guided Soft Signals · 2026

  • To explore the application of the method to other domains. To investigate the robustness of the method to different types of attacks. To develop more advanced methods for capturing dispersed attribution signals.

    Who Generated This 3D Asset? Learning Source Attribution for Generative 3D Models · 2026

  • Prior work has not systematically studied passive source attribution for generated 3D assets. The problem faces challenges such as dispersed attribution signals and realistic deployment constraints.

    Who Generated This 3D Asset? Learning Source Attribution for Generative 3D Models · 2026

  • The experiments were run in a simulated environment using the Habitat simulator. The evaluation was limited to six apartment scenes from the ReplicaCAD dataset.

    Fixed External Cameras as Common Prior Maps for Active 3D Scene Graph Generation · 2026

  • Extending the proposed method to more complex environments and scenarios. Investigating the use of multiple external cameras as Common Prior Maps. Improving the efficiency of active exploration in various applications.

    Fixed External Cameras as Common Prior Maps for Active 3D Scene Graph Generation · 2026

  • The datasets used in this work do not provide ground-truth relational annotations. The system relies on a deterministic geometry-based module for edge generation. The experiments are limited to the Replica and ReplicaCAD datasets.

    RGB-only Active 3D Scene Graph Generation for Indoor Mobile Robots · 2026

  • Current approaches to 3D scene graph generation rely on dedicated depth sensors, limiting deployment to specialized robotic platforms. Traditional perception pipelines lack semantic context or relational structure.

    RGB-only Active 3D Scene Graph Generation for Indoor Mobile Robots · 2026

  • The method encounters difficulties at very high magnifications (e.g., ×1024), where current vision-language models struggle to infer coherent structures. The system may not perform well with low-quality input images or limited training data.

    GaussianZoom: Progressive Zoom-in Generative 3D Gaussian Splatting with Geometric and Semantic Guidance · 2026

  • Investigating more capable content creative zoom-in approaches to enable seamless transitions from cosmic-scale environments down to microscopic and molecular scenes. Improving the performance of the system at very high magnifications (e.g., ×1024). Exploring the application of GaussianZoom to various fields such as architecture, product design, and video games.

    GaussianZoom: Progressive Zoom-in Generative 3D Gaussian Splatting with Geometric and Semantic Guidance · 2026

  • One challenge is the lack of a robust background matting framework for Virtual Production. Another challenge is the need to preserve pretrained semantics and improve robustness to background shifts.

    CineMatte: Background Matting for Virtual Production and Beyond · 2026

  • The research gap is the lack of a robust background matting framework for Virtual Production. Prior work has limitations, such as boundary artifacts and overfitting.

    CineMatte: Background Matting for Virtual Production and Beyond · 2026

Most-cited papers in Advanced Vision and Imaging

Most recent work

Find a gap in your own Advanced Vision and Imaging sub-topic

This page shows what the Advanced Vision and Imaging literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Computer Science

Online Learning and Analytics
1,106 open questions
AI in Service Interactions
861 open questions
Quantum Computing Algorithms and Architecture
645 open questions
Teaching and Learning Programming
623 open questions
Digital literacy in education
617 open questions
Music Technology and Sound Studies
576 open questions
Virtual Reality Applications and Impacts
496 open questions
Topic Modeling
476 open questions
122 open questions have been extracted from the limitations and future-work passages of 533 Advanced Vision and Imaging papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.