Computer Vision - Weekly Roundup

Generative design can now produce moldable parts by transforming 3D shape generation into 2D paired depth image generation, bypassing undercuts by construction.

ActionRoPE achieves geometrically consistent control in planar-camera games by mapping actions to world coordinates, eliminating the need for models to learn motion from scratch and reducing revisit drift by 66%.

Generative deep learning can synthesize realistic space radiation error patterns for NVM, achieving 92% fidelity with observed bit flips and capturing 91% of observed errors.

State-of-the-art text-to-image generation is achievable with 18x less compute by training from scratch, bypassing expensive language pretraining.

Robust prediction from corrupted latents, not diffusion objectives, unlocks superior visual tokenizers for diffusion models.

Generative design can now produce moldable parts by transforming 3D shape generation into 2D paired depth image generation, bypassing undercuts by construction.

ActionRoPE achieves geometrically consistent control in planar-camera games by mapping actions to world coordinates, eliminating the need for models to learn motion from scratch and reducing revisit drift by 66%.

Generative deep learning can synthesize realistic space radiation error patterns for NVM, achieving 92% fidelity with observed bit flips and capturing 91% of observed errors.

State-of-the-art text-to-image generation is achievable with 18x less compute by training from scratch, bypassing expensive language pretraining.

Robust prediction from corrupted latents, not diffusion objectives, unlocks superior visual tokenizers for diffusion models.

A novel source-anchored flow model achieves state-of-the-art video overlay removal, outperforming prior methods by over 70% on key perceptual metrics.

Human egocentric recordings can be automatically transformed into rich world models, unlocking scalable learning of spatial and temporal reasoning without extensive manual annotation.

Most action-conditioned video models struggle with hidden state, but a novel framework achieves 92% rule compliance by separating state evolution from pixel generation.

Egocentric video reconstruction is now metric and coherent, with a single model jointly recovering camera, scene, hands, and objects in a shared world frame.

Querying just 40% of trajectories with entropy gating unlocks 99% of the accuracy gains from full map context in travel-mode recognition.

Neural networks for point clouds require quadratic representation orders to guarantee separation, a concrete bound finally established.

Explicitly modeling ego-relative motion, spatial configuration, and TTC risk as first-class graph properties unlocks precise vehicle intention prediction from minimal ego-centric driving data.

Dynamically routing multiple frozen visual experts based on language context unlocks new state-of-the-art in weakly supervised referring expression comprehension.

Sparse attention in video diffusion models can be corrected using cached dense refinements, recovering fidelity without altering the reuse schedule.

Adversarial attacks create provable geometric inconsistencies across prime-quantized and CNN embedding spaces, enabling theoretically guaranteed detection.

A novel agentic framework achieves high-fidelity manga animation by grounding generation in source panels through iterative script and storyboard refinement.

Human-crafted conference posters are now directly usable as verified supervision for generating new posters, bypassing the limitations of layout-box-only datasets.

LLMs can now "think with visual latents" by generating masks directly from their internal representations, dramatically improving referring comprehension and segmentation.

Agents can now learn to design and critique 3D scenes like human designers, achieving state-of-the-art quality through complementary learning.

Packed trajectories, not sequential rollouts, unlock high-fidelity multi-turn image editing by preventing visual drift.

Open-weight diffusion models can now render complex, bilingual text and control intricate layouts with near-perfect accuracy, closing a critical gap with proprietary systems.

Video world models fail at executing instructed changes, not remembering them, a bottleneck MIST and M3-Agent finally expose and overcome.

A single scene can be represented by ~100K plane primitives (5MB), enabling video world models to maintain geometric consistency across hundreds of frames and large viewpoint changes.

Explicitly grounding the goal object in view unlocks near-expert performance in object-goal navigation, revealing perception rather than policy as the key limitation.

Explicitly aligning video captions with atomic audiovisual events unlocks substantial gains in fine-grained temporal and attributional accuracy, outperforming unstructured generation by over 9% on event-centric metrics.

A novel event-centric evaluation framework, AVC-Reward, finally provides a nuanced assessment of audio-video generation beyond holistic metrics, revealing fine-grained cross-modal failures.

Generative time series models can now produce multiple plausible future trajectories as fast as quantile models query a single point, unlocking accurate dependency modeling for critical applications.

Audio embeddings capture physical damping properties far better than single visual frames, and this gap is partially bridgeable through audio-to-vision distillation.

Diffusion models can learn new objectives by stochastically routing attention within their existing architecture, achieving 4x faster RL training without sacrificing generation quality.

Fine-tuning wind-field super-resolution models with persistent homology and merge tree losses dramatically improves topological structure reconstruction, outperforming standard methods and even the pretrained model.

A single model query can now estimate the full acceptance probability for parallel masked diffusion, slashing rejection sampling costs.

Wireless signals, unhindered by darkness or walls, can reconstruct building layouts by leveraging a diffusion model over physics-grounded representations of channel measurements.

AI weather models now forecast tropical cyclone tracks with superior probabilistic accuracy to leading physics-based systems, even while struggling with intensity.

Standard MIL suppresses secondary findings in coexisting lesions; this novel framework recovers them by modeling the dominant-label annotation process.

A single INR can now model complex, periodic deformations across entire image sequences, enabling self-regularization and outperforming prior methods.

A single diffusion model can forecast sky images and irradiance simultaneously, recovering ~46 W/m² of predictive power lost by cascaded frame-to-irradiance methods.

Bidirectional future graphs allow later evidence to retroactively correct earlier motion prediction segments, achieving state-of-the-art forecasting accuracy without LiDAR or ensembling.

Streaming VLMs can now internalize long-term history as evolving LoRA weights, bypassing context window limitations and achieving superior performance on video understanding tasks.

Generative models can inject crucial scientific knowledge into GP posteriors, recovering hidden modes and improving predictive accuracy where traditional GPs fail.

Hyperspectral image denoising and destriping now achieve state-of-the-art accuracy without any parameter tuning, outperforming methods that require true noise levels and running orders of magnitude faster.

Generative models can achieve significantly higher output diversity and corpus alignment by decoupling sampling from rendering at test time.

Video generators can be directly transformed into powerful 4D perception systems without iterative refinement, achieving SOTA on tracking and pose estimation.

Achieve "What You See Is What You Get" fidelity in single-image 3D texturing by decoupling material estimation and using observation latents as fixed anchors.

Adaptive uncertainty gating and post-crossing correction boost multi-object tracking accuracy by +1.0 HOTA on sports datasets, outperforming existing methods at real-time speeds.

Visual generation supervision injects significant, yet often inaccessible, knowledge into UMMs, requiring novel alignment strategies to unlock its full potential for text-based reasoning.

Small, fast-moving objects and fine textures in dynamic volumetric video are finally recoverable by intelligently allocating Gaussians based on future motion, not just current positions.

Predict future 3D scenes without needing to track camera pose, enabling consistent scale and cross-view understanding.

A simple, fixed query rule improves FID by over 2 points on CIFAR-10 without retraining, revealing prediction horizons are state-sensitive inference controls.

Most AI models miss subtle Chinese character rendering errors, but a new verifier using Ideographic Description Sequences spots them with 84% F1, outperforming OCR and VLMs by a wide margin.

A compact CNN achieves over 90% accuracy in real-world potato disease detection, outperforming larger models in efficiency for smallholder farmer applications.

Driving models can achieve near-perfect performance with just one denoising step by learning to predict visual motion directly, outperforming multi-step prediction models and achieving 3.29x inference speed.

Multimodal agents can now acquire evidence with verifiable optimality guarantees, reducing costly interactions by 90% while maintaining high decision accuracy.

A neural operator reconstructs continuous HRTF fields from sparse measurements with 18% lower LSD and 43% lower ILD error than leading baselines, achieving 373x speedup over neural fields.

Asynchronous EEG event detection is solved by a novel multiscale discrete spatiotemporal tokenization that captures both brief and sustained neural dynamics, outperforming prior methods across diverse datasets.

A simple reconstruction pre-training stage unlocks state-of-the-art one-step generative modeling with 10x less compute.

Spectral convolutional kernels unlock significantly greater accuracy gains from learned resolution allocation than their spatial counterparts, even when spatial area is normalized.

Deep learning DoA estimators can drastically cut inference time, but SVD-based beamforming remains the primary latency bottleneck in real-time ISAC.

Actively repelling uncertain features during representation learning slashes extreme segmentation errors in ambiguous anatomical boundaries by 22x, even without sacrificing overall accuracy.

A single turbulent flow realization, when intelligently patched, can unlock extreme-scale super-resolution, outperforming models trained on multiple full snapshots.

Continual pretraining a ViT encoder with LoRA on prostate MRI segmentation significantly improves downstream task alignment and performance, particularly for segmentation tasks, by inducing targeted representational shifts.

One-step RF signal separation achieves superior BER and MSE by learning a direct clean-endpoint transport via rectified flow matching.

Poisson noise, not just Gaussian, unlocks powerful generative models for a critical class of inverse problems.

Few-shot behavior cloning for autonomous driving achieves robust traffic light compliance by adaptively weighting demonstrations based on action density.

Prioritizing semantic integrity in early generative stages unlocks superior reward-mode balance in flow matching models.

Synthetic experience unlocks substantial game control improvements for LLMs, pushing success rates on complex benchmarks like MineDojo-300 from 12.63% to over 38%.

Real-time video object removal is now possible at 26 FPS, enabling interactive editing of arbitrarily long videos without latency or quality degradation.

Standard OCR metrics treat a typo and a wrong name equally, but LLM judges reveal severity differences that humans care about, leading to better evaluation.

Generative CT super-resolution can now reconstruct anatomically accurate vascular networks by explicitly guiding diffusion models with refined 3D vascular structures.

Frontier multimodal models are beginning to predict future world states, but current benchmarks reveal vast performance disparities and highlight the critical role of underlying language model capabilities.

Interactive video generation can now achieve high quality with large context windows without sacrificing real-time responsiveness.

Diffusion models can learn lattice field theories without ever seeing real data, matching HMC performance.

A single dimensionless motion prior generalizes across vastly different object families and data modalities, outperforming specialized methods without retraining.

A single latent cache can unify diverse visual representations and spatiotemporal queries for efficient, consistent dynamic world modeling across multiple video generation tasks.

Generate graphs that *exactly* realize a target persistence diagram, moving beyond approximate topological matching.

A single MLLM can now perform temporal grounding and video QA with significant inference speedups by treating them as structured decision tasks.

LLMs can now actively seek and ground verifiable spatio-temporal evidence, including object bounding boxes, to improve video reasoning accuracy by over 12%.

Generative 3D reconstruction now scales to entire indoor scenes, producing coherent, unified meshes with state-of-the-art fidelity from unposed images.

An agent can perform city-scale 3D object grounding by actively exploring a multi-view environment, bypassing the need for precomputed proposals and achieving state-of-the-art results.

Aesthetic code generation is unlocked by using a VLM judge to translate visual design principles from a reference image into actionable rewards for code models.

Humanoid robots can now learn complex tasks from egocentric, multimodal human demonstrations thanks to a wearable observatory and a novel motion reconstruction pipeline that achieves state-of-the-art accuracy.

Image tokens can be compressed by 94% by prioritizing semantic detail over spatial uniformity, unlocking efficient vision-language models.

Fine-grained, foundation-model-derived semantic representations unlock state-of-the-art whole-body PET/CT registration without task-specific fine-tuning.

VLMs can be taught to *use* latent geometry by training them to read continuous geometric tokens, rather than just reconstructing geometry from RGB.

Training-free conditional sampling for Schrödinger Bridges is now possible, achieving state-of-the-art results on image generation and inverse problems by twisting SB's inherent endpoint prediction capabilities.

Mixing the right local appearances, not just blending pixels, unlocks significantly higher quality texture interpolations.

A single dataset now unifies 279,166 multi-sensor, multi-temporal Earth observation tiles, unlocking efficient pretraining for generalist foundation models.

Physical controls like damping and size can precisely guide audio foley generation, achieving near-perfect correlation with real-world sound targets without fine-tuning.

Segmentation models implicitly learn powerful object-centric representations that outperform encoder-based feature extraction for medical image prediction, even without explicit downstream supervision.

Robust generative modeling is achieved by replacing squared error with geometric median regression in Flow Matching, outperforming standard methods under severe data contamination.

Physically informed GNNs can accurately forecast advection-diffusion dynamics on irregular grids, outperforming existing methods and enhancing solar power predictions.

Data replay methods can recover up to 88% of performance in task-incremental radar waveform classification, with synergistic combinations of replay mechanisms boosting accuracy by over 5%.

Two residual streams in diffusion language models halve the performance gains of adding a second stream by enabling distinct read and write routing patterns.

Most industrial anomaly detectors collapse under simple brightness or zoom changes, losing up to 34 AUROC points, but a new approach using synthetic anomalies maintains over 92% AUROC.

Generative image and video compression achieves significant memory and compute reduction by replacing dense Gaussian codebooks with implicit randomized Hadamard dictionaries, enabling real-time streaming video compression on a single GPU.

Few-step diffusion models can achieve state-of-the-art 4-NFE generation without auxiliary score models or adversarial training, matching or exceeding multi-step teachers.