Efficient ViTs

Comparing transformer backbones for image segmentation is confounded: each is paired with a different decoder, recipe, and pretraining, so reported differences rarely reflect the backbone itself. We introduce the Lightweight Universal Mask Adapter (LUMA), a lightweight, backbone-agnostic mask-transformer head that treats any backbone as a black-box feature extractor, letting a set of queries read from its features through cheap cross-attention. LUMA matches the accuracy of EoMT, the state-of-the-art efficient ViT-segmenter, at lower cost, while attaching unchanged to isotropic, hierarchical, convolutional, and mixture-of-experts backbones alike. Holding this head fixed, we benchmark 20 backbones, 11 pretraining schemes and a range of resolutions on ADE20K and Cityscapes under one modern recipe. We find that ``efficient'' token mixers fail to deliver efficiency even at the high resolutions that motivate them, with plain ViT holding the throughput Pareto-front at every resolution. Additionally, the pretraining objective, not the architecture, the lever the field has tuned hardest, governs segmentation quality.

Deploying Vision Transformer (ViT) models on edge platforms remains challenging due to their high computational demands and the architectural heterogeneity of modern hybrid ViT models, which incorporate both fully connected and convolutional layers. This heterogeneity leads to significant variation in tensor shapes, requiring flexible and efficient FPGA-based acceleration. In this paper, we present FlexViT, a reconfigurable FPGA accelerator for efficient ViT inference on resource-constrained edge devices. Built on the SECDA-TFLite framework, FlexViT employs a hardware-software co-design approach that maps both fully connected and convolutional layers onto a unified high-throughput INT8 GEMM engine using a runtime im2col transformation. To efficiently support diverse layer configurations, we propose a dual-mode dataflow that dynamically switches between input and weight reuse by reconfiguring the compute array at runtime. We further introduce a depth-first tiling strategy that completes accumulation in a single pass, eliminating off-chip partial-sum transfers and reducing memory bandwidth requirements. We implement FlexViT on a PYNQ-Z2 FPGA and evaluate it across a representative set of ViT models. FlexViT achieves up to 2.74x speedup on accelerator-executed layers, translating into up to 1.40x end-to-end speedup compared to CPU-only execution. The code is available at: https://github.com/gicLAB/FlexViT

Most token reduction methods for Vision Transformers seek favorable tradeoffs between accuracy and efficiency by pruning, merging, or pooling patch tokens. REDI (Relevance for DINOv3 Token Reduction) studies this question through a controlled supervised reference: how should a fixed token budget be allocated across patches for image classification? REDI quantizes final block DINOv3 patch representations into a visual vocabulary and derives class conditioned corpus scores using supervised TF-IDF over visual words. For each validation image, the ground truth class selects a row of the TF-IDF table, and four transformed views produce a TF-IDF map aligned to a reference center crop. A separate dense pass on the same crop provides an attention map. After independent min max normalization, their elementwise product defines the REDI score. A fixed keep, merge, and compress operator then uses score rank to assign patch roles and score magnitude to weight merging and compression. With precomputed REDI scores, a frozen DINOv3 ViT-B/16 backbone, and the same linear classifier used for dense evaluation, the operator reduces the sequence length from 201 to 107 tokens, a 46.8% sequence reduction. The REDI variant based on incoming attention mass achieves 84.706% Top-1 accuracy on ImageNet-1K, compared with 83.514% for the dense baseline, 82.634% for incoming attention mass alone, and 81.796% for supervised TF-IDF alone. The same corpus term also improves reduced classification for three alternative attention formulations relative to their attention only counterparts. Together, these controlled comparisons indicate that class specific corpus statistics and image specific attention provide complementary signals for patch ranking in this setting.

We present RenderFormer++, a scalable and physically grounded feed-forward neural rendering framework for global illumination in mesh scenes. Existing Transformer-based neural rendering methods such as RenderFormer achieve promising cross-scene generalization, but suffer from limited physical consistency and poor scalability due to the quadratic attention complexity of triangle-level tokenization. To address these issues, we introduce Physics-Informed Transport Guidance (PITG), which embeds rendering-equation inductive biases into the attention mechanism and enforces transport consistency loss, enabling physically consistent light transport modeling. We further propose Hierarchical Object-Centric Tokenization (HOCT), which aggregates triangle-level features into compact object-level tokens via cross-attention with learnable queries, substantially reducing computational and memory costs while preserving geometric and radiometric information. Extensive experiments demonstrate that RenderFormer++ achieves scalable, stable, and generalizable feed-forward global illumination rendering across complex large-scale scenes with improved physical accuracy and efficiency over prior neural rendering methods.

Modern VLMs and VLA systems commonly adopt off-the-shelf ViTs such as SigLIP2 as visual encoders, but diverse downstream requirements in latency, temporal modeling, and VLM integration often call for customized SOTA-level ViTs. Training such encoders remains beyond the reach of much of the community, as it requires massive image-text data, while standard softmax attention makes high-resolution or dynamic-resolution pretraining prohibitively costly and often forces low-resolution pretraining followed by post-hoc adaptation. TuringViT addresses these challenges with three key designs: Turing Linear Attention (TLA) for efficient sequence modeling, VISTA-Curation to construct supervision-rich image-video training data, and native dynamic-resolution pretraining that supports flexible inputs from the start and transfers seamlessly to downstream VLMs. As a result, TuringViT outperforms leading open-source ViT baselines with only 10% of the data, achieves stronger downstream VLM performance, and delivers substantially better latency scaling on high-resolution inputs. Our scaling-law analysis further shows that TuringViT continues to improve predictably with curated data scale, far from saturation. Its fast adaptation, hardware-friendly design, and efficient deployment have made it a unified visual foundation across XPeng's AI systems. More broadly, TuringViT provides a reproducible pipeline that dramatically lowers the cost for the community to train, customize, and deploy SOTA-level ViTs, moving toward making such Vision Transformers accessible to all.

Deploying vision transformers (ViTs) on sensor-edge systems is limited not only by on-device compute, but also by the energy and bandwidth required to transmit high-dimensional image data from the sensor to the processor. While in-sensor and near-sensor computing reduce this cost through early feature extraction, existing methods often provide only modest compression. We observe that the frequency domain provides a naturally compact representation of visual information and can be exploited at the sensor level to reduce sensor-to-processor data movement. Building on this insight, we present FrequencyFormer, a co-designed sensor-to-processor pipeline for efficient ViT inference. FrequencyFormer includes: (1) a multi-scale DCT tokenizer that compresses a 224x224 image into compact frequency-domain tokens, achieving up to 128x reduction in off-chip data volume with modest accuracy loss; (2) a LUT-based near-sensor hardware implementation that leverages fixed DCT coefficients for multiplier-free, energy- and area-efficient tokenization; and (3) a modified MIPI-based low-power communication architecture that further reduces transfer energy. FrequencyFormer serves as a drop-in replacement for standard ViT patch embedding and remains compatible with pretrained backbones across classification, detection, and segmentation tasks. The pipeline achieves 28.8 TOPS/W, reduces communication energy by 230x, and lowers total sensor-side energy by 2.22x, demonstrating frequency-domain tokenization as a scalable foundation for in-sensor ViT deployment.

Vision Foundation Models (VFMs) with Vision Transformer (ViT) backbones, such as DINOv2, have become essential for downstream tasks like object recognition and semantic segmentation. The immense computational requirements of backbones often necessitate distillation into smaller architectures for edge deployment. Feature-based knowledge distillation (KD) often suffers from the teacher-student gap; the student struggles to imitate teacher's complex feature map due to its limited capacity. To mitigate this bottleneck, we propose Dyna-DINO: Efficient ViT Distillation Via Adaptive Representation Anchoring, a training curriculum for ViT feature-based knowledge distillation. By utilizing the teacher's intermediate feature maps as a sequence of progressively more difficult targets, our curriculum allows the student to build a foundational representation before tackling higher-level abstractions. Our results demonstrate that this paradigm significantly accelerates convergence through adaptive difficulty selection across various student model sizes and dataset scales. With our curriculum, the Dyna-DINO distilled ViT-S achieves 90.1% accuracy on ImageNet-100, a +12.24% improvement compared with baseline. On ImageNet-1K, Dyna-DINO achieves +3.9% and +6.09% improvement for the instance retrieval task on the Oxford and Paris datasets, +1.93% improvements on semantic segmentation task, as well as meaningful performance gain on classification task. Furthermore, the curriculum enables 25.1% savings in training FLOPs and 21% savings in training time on ImageNet-100 by implementing early-stopping for teacher inference during the initial stages of training. Code is available at https://github.com/KevinZ0217/Dyna-DINO

Visual Geometry Grounded Transformer (VGGT) recovers dense 3D scene structure from multi-view images in one forward pass, but quadratic cross-frame attention limits its scalability. Existing training-free accelerators reduce computation uniformly along one axis, missing layer heterogeneity. Our spectral, probing, and causal analyses reveal three regimes: shallow layers lack cross-view structure, middle layers drive cross-view alignment, and deep layers are redundant for dense geometry yet their cross-frame attention remains essential for pose. RegimeVGGT applies layer-wise U-shaped compression along two axes: Saliency-Guided Banded Merging protects geometry- and edge-salient tokens, while Selectively Protected K/V Downsampling preserves cross-frame spatial coverage and the pose-critical path through a phase-shifted spatial grid, a reference-frame anchor, and uncompressed camera/register tokens. Training-free, RegimeVGGT achieves a 6.7x speedup over VGGT* at matched reconstruction quality.

Text-guided open-vocabulary object counting (TOOC) aims to count objects belonging to the categories specified by natural language descriptions. Although vision-language pre-trained models have been successful applied to TOOC tasks, they still struggle with fine-grained spatial understanding and real-time inference requirements in counting scenarios. To address these limitations, this paper proposes a real-time TOOC framework, called the Real-Time Counter (RT-Counter), that achieves not only good counting accuracy but also high computational efficiency. RT-Counter designs a novel Visual Prototype Textualization (VPT) module that can project learned visual features into a text feature space and then generate features containing the abstract information that is hard to capture with visual prototypes and the detailed prototype information that is difficult to describe in text, enhancing the object-level visual-language model's counting capabilities. Additionally, RT-Counter incorporates our Weaving Transformer (Weaformer) layers, maintaining high descriptive power at a fraction of the computational cost. The Weaformer layer adopts a novel hybrid attention mechanism that can efficiently weave together local and global visual features. Extensive experiments on three public datasets show that RT-Counter successfully breaks the accuracy-speed trade-off in TOOC. While achieving a competitive MAE of 13.30 on FSC147, RT-Counter operates at 112.48 FPS, making it 7.4x faster and over 4$\times$ more parameter-efficient than the existing leading methods in TOOC. Our work aims at balancing high accuracy and real-time performance in TOOC. Code is available at: https://github.com/Jason-Mar1/RT-Counter.

Industrial inspection needs zero-shot anomaly detection (ZSAD) that remains useful under edge deployment constraints. Recent methods often rely on ViT-L foundation backbones (~300M parameters), which exceed the memory and operator budget of typical embedded hardware. We study this regime through EdgeZSAD, a compact reference system built around a TinyViT-21M-512 backbone, an asymmetric global-local readout (EdgeGLR), and a reproducible source-side training recipe (Real-IAD-DR). We train a single checkpoint in a source-trained, target-unseen protocol and evaluate it across six industrial benchmarks. Across three independent runs, the resulting model reaches an average image AUROC of 91.6 on MVTec-AD and 88.2 on VisA, while remaining directly deployable on Jetson Orin Nano Super (TensorRT FP16) and RB5 Gen2 (QNN GPU FP16). Across the six device-rescored benchmarks, image-AUROC drift stays below 0.2 points, indicating that the exported graph preserves host-side ranking behavior in the evaluated deployment setting.

Vision Transformers (ViTs) have gained significant attention in computer vision and shown strong potential for face recognition (FR). However, their high computational cost makes deployment on resource-constrained devices challenging, motivating the need for methods that balance efficiency and accuracy. In this work, we investigate early exiting in pretrained ViTs as a simple yet effective training-free strategy for efficient FR inference. Leveraging the uniform feature dimensionality across transformer encoder blocks, we introduce ViT-FREE, a multi-exit framework that enables face verification directly from intermediate representations without modifying or retraining the backbone model, and thus, reducing inference cost. Empirically, we show that patch embeddings and attention maps evolve progressively across depth, exhibiting high similarity between consecutive ViT blocks and increasing alignment with the final representation. This indicates gradual feature refinement and attention convergence, suggesting that intermediate layers already provide stable and discriminative representations suitable for early exiting. Through extensive experiments on multiple FR benchmarks, we systematically analyze the accuracy-efficiency trade-off across exit depths. Our results demonstrate that later exits achieve a highly favorable balance, with exiting at layer 10 yielding up to a 20% speedup while incurring only a 1.5 drop in verification performance on benchmarks such as IJB-C. Also, we propose ViT-FREE_FT, a lightweight exit-specific fine-tuning strategy that adapts only the projection layers using a small synthetic dataset while keeping the transformer backbone frozen. This approach improves the performance of shallow exits while preserving the efficiency benefits and leaving deeper exits largely unaffected.

Though Vision Transformers (ViTs) have become the dominant backbone in many computer vision tasks, due to permutation equivariance, their attention mechanism lacks explicit spatial inductive biases. This become particularly important in two settings: when model capacity is small or training data is limited. Inspired by the attention masking strategies in Linear Transformers and the scanning patterns of Vision SSMs, we introduce VIOLIN, a lightweight masked attention mechanism that encodes spatial structure within attention via Space Filling Curves (SFCs) with less than 0.0015% extra parameters and negligible computational overhead. VIOLIN scans the image using multiple SFCs to construct curve-specific decay masks, which are then combined and multiplied with the attention matrix. Across a wide range of evaluations, VIOLIN consistently improves performance. In limited data regimes such as fine-tuning on VTAB-1K, it boosts accuracy across all task groups and by up to 8.7% on the tasks where spatial information is essential. It can be combined with parameter-efficient fine-tuning methods such as LoRA to further increase the performance. Beyond fine-tuning, VIOLIN improves various small scale ViT architectures (e.g., DeiT, DINO) during pretraining on ImageNet-1K. Additionally, on pixel-level CIFAR-100 training, a task that is highly dependent on location information, VIOLIN increases accuracy by up to 7.2%. Overall, VIOLIN provides a computationally efficient yet effective way to inject spatial inductive bias into ViTs, especially benefiting small models and limited data settings.

Image captioning is a challenging and significant task that aims to generate coherent and semantically meaningful textual descriptions for given images. To accomplish this task, it requires a deep understanding of visual content along with the ability to express that understanding in natural language. Despite remarkable progress with transformer-based architectures, existing approaches often suffer from limitations, such as a lack of rich local feature representations and the high computational cost of quadratic self-attention. The proposed model focuses on improving computational efficiency by restructuring the vision transformer architecture. In designing this approach, the standard self-attention mechanism in Vision Transformers is replaced with a probabilistic transformer approach based on a Gaussian Mixture Model (GMM), a soft-clustering technique. Instead of computing pairwise attention among all image patches, the model groups similar patches into a fixed number of clusters using an Expectation-Maximization (EM) algorithm. This clustering-based mechanism reduces the computational complexity from quadratic O(n^2) to linear O(nK), where K << n. The autoregressive GPT-based decoder is used for caption generation. The model is evaluated on the Flickr 30K dataset, demonstrating competitive and significant improvement over existing works.

Remote sensing applications for environmental monitoring and disaster management are frequently constrained by a spatial--temporal trade-off: imagery with fine spatial detail is often acquired less frequently, whereas more temporally available observations are typically coarser. Single-image super-resolution provides a practical means to enhance coarse imagery without changing acquisition schedules, yet many Transformer-based SR models remain computationally expensive and can be sensitive to limited or geographically biased training data, which degrades robustness under out-of-distribution conditions. This paper presents NGram-MoSE, a lightweight Transformer architecture designed to improve both efficiency and texture continuity. NGram-MoSE introduces N-Gram Context Injection to strengthen cross-window local consistency and mitigate window-boundary artifacts, and incorporates a Mixture-of-Experts (MoE) feed-forward design to scale capacity through sparse activation without proportional growth in inference cost. Experiments on a geographically disjoint OOD test set show that NGram-MoSE achieves 31.68,dB PSNR while reducing FLOPs by $14\times$ relative to a heavyweight Transformer reference. Downstream evaluation on a landslide segmentation benchmark further demonstrates that restoring degraded inputs to the detector training scale improves performance, yielding a 4.47% absolute gain in mAP@50 over bicubic upsampling, and exhibits stronger cross-scale consistency under scale extrapolation. These results indicate that NGram-MoSE provides an effective SR module for resource-constrained remote sensing pipelines requiring robust generalization.

Vision Transformers (ViTs) achieve strong performance but suffer from high computational costs due to quadratic self-attention complexity. Although token reduction techniques such as pruning and merging mitigate this, they typically overlook how representations evolve across network depth. We propose RAPID, a depth-aware token reduction framework that adapts reduction strategies to the layer-wise characteristics of token representations. The primary methodological contribution is a bifurcated strategy: in shallow-to-middle layers, RAPID employs a redundancy-similarity aware pruning metric to eliminate over-represented local patterns. As features transition to global semantic concepts in deeper layers, the framework shifts to an importance-similarity aware merging mechanism. This stage leverages classification (CLS) token attention weights to protect semantically critical tokens while fusing less important but similar neighbors. Empirical validation on ImageNet-1K using ViT and DeiT architectures demonstrates that RAPID establishes a superior accuracy-compression Pareto frontier compared to plug-and-play baselines such as ToMe and ToFu. RAPID is particularly robust in aggressive compression regimes, achieving up to 4.29% higher accuracy than ToMe at extreme reduction rates. Our framework provides a training-free template for optimizing vision models by aligning reduction strategies with hierarchical feature evolution.

Cloud removal aims to accurately reconstruct the ground objects obscured by clouds in remote sensing images. Existing Transformer-based methods utilizing self-attention have shown impressive results by effectively modeling long-range dependencies in cloudy images. However, they suffer from the following issues: 1) the high computational complexity of self-attention limits scalability; 2) treating both cloudy and clean pixels as valid within the attention computation brings disturbances in subsequent layers, leading to suboptimal performance. To address these challenges, we propose the Adaptive Triangular Transformer for Cloud Removal (ATT-CR), a model that effectively reduces computational costs and mitigates interference from cloudy pixels. Specifically, it consists of two core components: Triangular Attention (TAN) and Feature Selected Gating Module (FSGM). TAN employs lower and upper triangular matrices to approximate Softmax attention with O(N) computational complexity, significantly reducing the computational costs. The FSGM, on the other hand, integrates with TAN to adaptively distinguish between cloudy and clean features, which minimizes the introduction of invalid information into subsequent layers. Extensive experiments on cloud removal benchmarks demonstrate that ATT-CR delivers superior performance compared to existing methods.

Vision Transformers (ViTs) have demonstrated strong representation capability in image classification. However, their quadratic self-attention complexity and large parameter counts limit deployment on resource-constrained mobile and edge devices. This paper introduces UtVAA, an ultra-tiny Vision Transformer architecture designed for efficient visual recognition under strict computational budgets. It incorporates a novel Affix Attention block that combines depthwise-pointwise local feature extraction, linear self-attention, coordinate attention for spatial dependency modelling, and a lightweight ternary fusion strategy to integrate local and global representations. In addition, Dilated Bottleneck blocks expand the receptive field using dilated depthwise separable convolutions while maintaining low FLOPs and stable optimisation through residual connections. UtVAA is implemented in scalable Tiny, Medium, and Large variants, with the smallest model containing 204.67K parameters and 53.95M FLOPs. Experimental results on CIFAR-10, CIFAR-100, PlantVillage-Tomato and SLIF-Tomato datasets show that UtVAA achieves competitive accuracy within a sub-million-parameter regime. Overall, the results demonstrate that transformer-based vision models can be redesigned into ultra-tiny architectures without significant loss in discriminative performance, making UtVAA suitable for mobile and edge deployment. Code is available at https://github.com/romiyal/UtVAA

Background: The lateral ventricle choroid plexus (LVCP) is gaining recognition as a key imaging biomarker for multiple sclerosis (MS) related to physical disability and neuroinflammation. Yet, manual segmentation of the LVCP is highly tedious, restricting its use in broad clinical trials and longitudinal assessments. This research aims to develop a SwinUNETR-driven pipeline that leverages targeted intra- and peri-ventricular small patch sampling to automatically segment the LVCP in MS from both standalone and multi-modal MRI inputs. Methods: We retrospectively assessed 3T MRI scans across three sets of data stemming from two separate MS-dominant cohorts (Dataset 1: n=177; Dataset 2: n=177; expanded test set: n=388). Our method employed a SwinUNETR architecture trained on 32x32x32 voxel patches, benchmarking it against the 3D UXNET model. The primary metric for evaluation was the Dice Similarity Coefficient (DSC), supplemented by computational demand (GFLOPs) and the 95th percentile Hausdorff Distance (HD95). Results: On the extended test set, the SwinUNETR model secured a mean DSC of 0.868 (95% CI: 0.863-0.872) with MPRAGE and FLAIR combined, showing a statistically significant gain over UXNET (DSC: 0.858 [95% CI: 0.853-0.862], p<0.0001). When restricted to standalone FLAIR inputs, the transformer-based approach sustained a high DSC of 0.863, while the spatial localization of UXNET worsened considerably (HD95: 1.86 vs. 3.00 mm). Importantly, the proposed framework lowered computational load by 99% (91.8 vs. 22,080 GFLOPs). By integrating localized patch sampling with a SwinUNETR architecture, this methodology offers an accurate, robust, and statistically superior alternative to current leading models for LVCP segmentation. Its vast reduction in computational cost makes it ideal for widespread implementation in clinical and research environments.

Spiking Vision Transformer (SViT) models are promising low-power ViT models for solving vision-based tasks with state-of-the-art performance. However, their large sizes limit their deployments for resource-constrained embedded platforms, underscoring the needs of model compression. One of prominent compression techniques is pruning, and the state-of-the-art works employ unstructured pruning techniques to compress SViT models. Such techniques require specialized hardware architectures tailored for the sparsity patterns to maximize their efficiency benefits, making this approach not scalable. To address this, we propose PSViT, a novel methodology to perform structured pruning on SViT models, hence making it possible to efficiently accelerate their inference using the existing and widely-used computing architectures. To do this, PSViT employs several key steps: uniform channel-wise filter pruning to structurally eliminate the non-significant weights, sensitivity analysis to evaluate the impact of channel-wise pruning of individual layer on accuracy and network size, as well as fine-grained channel-wise pruning based on the sensitivity analysis and the given network architecture. Experimental results show that PSViT effectively obtains 22.4% memory saving through single-shot pruning, while maintaining high accuracy within 3% (70.3% without fine-tuning and 72.8% with fine-tuning) from the original non-pruned SViT model (73.3%) on the ImageNet-1K. These results also show that the PSViT methodology advances the effort in enabling efficient SViT deployments on resource-constrained applications.

Semantic segmentation of remote sensing imagery requires models that capture both global context and local detail under tight computational budgets. Prior work typically optimizes for one of these axes: attention for global context, convolution for local detail, or compactness for efficiency. While hybrid approaches aim to capture both, they require architectural changes and encoder backbones with computational overhead, limiting efficiency and performance. We present LALE (Lightweight-transformer Architecture for Land-cover Estimation), an end-to-end remote sensing image segmentation architecture, that bifurcates its encoder by resolution: lightweight ConvMixer stages handle high-resolution local features, while transformer stages handle low-resolution global context, confining the quadratic cost of self-attention to deep, downsampled feature maps. An all-MLP multi-scale decoder, together with RMSNorm and StarReLU throughout, further reduces compute and parameter count. On the large-scale ARAS400k remote-sensing segmentation benchmark, LALE establishes a strong efficiency-performance trade-off against CNN, transformer, and hybrid baselines. Our smallest variant, (just 1.6M parameters), reaches within 2.6 F1 points of the best baseline (UPerNet) while using 4.5x fewer parameters, 7x less storage, 17x fewer GMACs, and delivering 1.8x higher throughput. The codebase for LALE is publicly available at https://github.com/caglarmert/LALE.

Compact Vision Transformers are attractive for medical imaging in low-data and resource-constrained settings, but most existing variants assume that Euclidean latent geometry is sufficient for organizing image representations. We introduce hZACH-ViT, a family of curved-geometry extensions of ZACH-ViT, a compact zero-token Vision Transformer that removes positional embeddings and the class token and relies on global average pooling over patch representations. To isolate the role of geometry, we preserve the verified ZACH-ViT backbone and modify only the final representation space and prototype-based classifier head, enabling a controlled comparison between Euclidean, hyperbolic, and spherical latent geometries. We evaluate Poincaré, Klein, and spherical hZACH-ViT heads on seven MedMNIST datasets under an identical few-shot protocol with 50 samples per class and five random seeds. The completed benchmark contains 770 training runs spanning seven datasets, three non-Euclidean geometries, seven curvature magnitudes, and a Euclidean baseline. Across all seven datasets, the best non-Euclidean hZACH-ViT configuration improves over Euclidean ZACH-ViT, with an average gain of +0.021 in the dataset-specific primary metric and the largest improvement on OCTMNIST (+0.055 MacroF1). Fixed low-curvature configurations retain positive gains on the majority of datasets, and low curvature values (c = 0.1 or 0.2) account for six of the seven dataset-level winners. Rather than identifying a universally optimal manifold, our results establish geometry and curvature as dataset-dependent model-selection variables, with fixed low-curvature analyses confirming that gains persist beyond exhaustive per-dataset tuning.

Vision foundation models are bottlenecked by the quadratic cost of self-attention, which limits usable resolution and increases the cost of large-scale pretraining. Subquadratic alternatives such as linear attention and state-space models reduce this cost, but often serialize images into 1D token streams and weaken the 2D spatial structure important for vision. Generalized Spatial Propagation Networks (GSPN) instead propagate context directly on the 2D grid through line-scan recurrences, achieving near-linear complexity without positional embeddings, but have seen little use as foundation-scale encoders. We present C-GSPN, a foundation-scale vision encoder based on 2D spatial propagation. C-GSPN makes the operator practical through three improvements: (1) a fast GSPN CUDA kernel that fuses per-step launches into a single warp-specialized implementation with shared-memory tiling, coalesced access, and a compact multi-channel propagation, reaching over 90% of peak memory bandwidth and running up to 40--52x faster than the original GSPN implementation; (2) a compressed latent-space propagation block with fused normalization, which turns kernel-level speed into block- and model-level efficiency; and (3) a two-stage cross-operator distillation recipe that trains the new architecture from an attention teacher without the cost of from-scratch foundation-scale training. Distilled with 600M image-text pairs, C-GSPN matches an isomorphic ViT baseline with 15% fewer parameters, improves ADE20K segmentation by +2.1%, transfers to high resolution with a fraction of the data needed from scratch, and delivers a 4x end-to-end block speedup at 2K with single-pass, tiling-free inference.

Real-world image deblurring demands both high-fidelity restoration and computational efficiency, a balance existing methods often struggle to achieve. In this paper, we propose FSM-Net (Frequency-Spatial Multi-branch Network), a highly efficient solution that secured 2nd place in the NTIRE 2026 Challenge on Efficient Real-World Deblurring. FSM-Net pioneers a dual-domain approach: a novel Frequency Attention module explicitly recovers high-frequency structural details via FFT, while a Cross-Gated Vision E-Branchformer at the bottleneck captures global dependencies with linear complexity. To ensure robust convergence, we employ a progressive curriculum training strategy guided by a composite loss function (Multi-Scale Charbonnier, Structural Edge, and Frequency). Evaluated on the RSBlur benchmark, FSM-Net achieves an outstanding 33.144 dB PSNR with only 4.94M parameters and 159.35 GMACs (at 1920x1200 resolution). By effectively pushing the Pareto frontier of efficiency and quality, FSM-Net establishes a strong baseline for resource-constrained image restoration.

Sparse vision transformers have gained popularity as efficient encoders for medical volumetric segmentation, with Swin emerging as a prominent choice. Swin uses local attention to reduce complexity and yields excellent performance for many tasks but still tends to overfit on small datasets. To mitigate this weakness, we propose a novel architecture that further enhances Swin's inductive bias by introducing Inception blocks in the feed-forward layers. The introduction of these multi-branch convolutions enables more direct reasoning over local, multi-scale features within the transformer block. We have also modified the decoder layers in order to capture finer details using fewer parameters. We demonstrate a performance improvement on eleven different medical datasets through extensive experimentation. We specifically showcase advancements over the previous state-of-the-art backbones on benchmark challenges like the Medical Segmentation Decathlon and Beyond the Cranial Vault. By showing that the existing inductive bias in Swin can be further improved, our work presents a promising avenue for enhancing the capabilities of sparse vision transformers for both medical and natural image segmentation tasks. Code and pre-trained weights can be accessed at https://github.com/Eiphodos/SwInception.

The recursive quad-tree partitioning in High Efficiency Video Coding (HEVC) incurs considerable computational overhead, with exhaustive rate-distortion optimization for CTU partition prediction consuming the dominant share of encoding time. Although partition prediction through deep learning has emerged as a viable encoding accelerator, an architectural dichotomy remains largely unaddressed: CNNs are computationally efficient but spatially myopic due to their localized effective receptive fields, failing to capture long range semantic relationships and repetitive textures; conversely, transformer based architectures are better at capturing global context but incur prohibitive CPU latency, a critical liability that impedes deployment which is predominantly CPU-bound. This paper introduces Hybrid Fast Vision Transformer (HFViT), a hybrid architecture designed to accelerate HEVC intra-mode partition prediction. HFViT fuses a reparameterized depthwise-separable convolutional backbone with a Hierarchical Attention Transformer (HAT) mechanism, leveraging a carrier token scheme to enable efficient global information propagation at sub-quadratic complexity. Post-training structural fusion collapses batch normalization into preceding layers to further reduce latency. Comprehensive evaluation reveals the efficacy of HFViT in accelerating HEVC intra-encoding across resolutions. On standard JCT-VC test sequences, HFViT reduces the average VMAF BD-rate penalty by 2.4, 2.6, and 7.9 percentage points on Classes A, B and E, respectively, as compared to the competing ETH-CNN baseline while maintaining CPU inference latency within 8% of the CNN baseline and surpassing it on GPU by 40%, establishing practical viability for real-time encoder integration.

Image semantic communication is a critical component in next-generation wireless communication systems. However, such systems typically suffer from large memory footprints and high computational complexity, making them difficult to deploy on resource-constrained devices. To address these challenges, we propose a vision transformer (ViT)-enabled image semantic communication system. In this system, a recursive structure is introduced to iteratively refine semantic features and reduce the parameter count. In addition, three dynamic adjustment strategies are designed to adaptively reduce computational complexity: dynamic depth adjustment, dynamic width adjustment, and joint width-depth optimization. Dynamic depth adjustment adaptively determines the number of recursive modules according to image content and channel conditions, while dynamic width adjustment selectively preserves important neurons and attention heads. The joint width-depth optimization further enables flexible computation configurations. Simulation results verify that the proposed recursive ViT-based system, combined with the three dynamic adjustment strategies, reduces the parameter count by 48.7% and achieves higher reconstruction quality than existing baselines under comparable computational complexity.

Given the real-time demands of UAV tracking, many methods simplify the backbone to reduce computation, but this often weakens feature representation and degrades performance in complex scenarios. To alleviate this issue, we propose EATrack, an efficient and asymmetric UAV tracking framework centered around a teacher-guided dual-branch distillation strategy that enhances the feature expressiveness of the lightweight student model. Specifically, EATrack investigates two complementary perspectives of knowledge transfer: spatially focused feature-level distillation that compensates for weakened representations by guiding the student to learn strong target representations, and prediction-level distillation that enhances spatial localization by learning the teacher's capability for accurate target localization. Furthermore, to enhance robustness against appearance variations, we introduce a fine-grained target-aware distillation strategy that selectively transfers the teacher's target modeling capacity to the student. A temporal adaptation module is incorporated at inference to enhance robustness over time. Experiments on five UAV benchmarks demonstrate that EATrack achieves a favorable balance between accuracy and speed. Code: https://github.com/GXNU-ZhongLab/EATrack

We introduce JetViT, a novel family of hybrid-architecture Vision Transformer (ViT) models that match the accuracy of state-of-the-art full-attention vision foundation models while achieving substantially higher inference efficiency on high-resolution images. At the core of our approach is Post-Training Attention Search, a post-training acceleration framework that converts pre-trained full-attention ViTs into efficient hybrid-attention variants by identifying and replacing redundant full-attention blocks with linear or window-attention blocks. By inheriting the MLP and attention weights from the base model, Post-Training Attention Search efficiently explores the architectural design space through three key steps: (1) optimizing the linear-attention block design; (2) finding the best combination of linear-attention and window-attention blocks; and (3) identifying and preserving critical full-attention blocks. We evaluate JetViT on two representative high-resolution vision foundation models, DINOv3 and DepthAnythingV2. On the NVIDIA H100 GPU, JetViT achieves up to 1.79x higher throughput and up to 44.81% lower latency without sacrificing accuracy. We will release our code and accelerated ViT models soon.

The quadratic cost of self-attention in Vision Transformers (ViTs) constitutes a fundamental bottleneck for practical deployment, motivating a vibrant line of research on token reduction. Among existing approaches, token merging (ToMe) has emerged as an elegant training-free solution; yet its design rests on an unspoken premise of token equality, which contravenes the well-documented non-uniformity of self-attention and leads to information loss in high-salience tokens under aggressive compression. We address this limitation with AdaMerge, a token-merging framework based on two complementary mechanisms. First, salience-weighted similarity leverages column-wise feature-affinity centrality as a token-importance proxy and incorporates the resulting salience scores into the bipartite matching score, ensuring that pivotal tokens contribute more strongly to the merged representation. Second, adaptive merging intensity uses pre-computed layer-wise similarity statistics to dynamically modulate the per-layer reduction count in accordance with input-specific redundancy. On ImageNet-1k with ViT-B/16, AdaMerge consistently outperforms ToMe, PiToMe, and DSM across all FLOPs-matched regimes. The accuracy gap widens monotonically with compression: at the 13.4G FLOPs operating point, AdaMerge sustains a Top-1 degradation of only -1.06%, compared to -1.45% for PiToMe and -4.62% for DSM. To our knowledge, AdaMerge is the first to combine salience-weighted similarity and adaptive per-layer reduction into a single training-free token merging framework, advancing the accuracy-FLOPs Pareto frontier of ViT acceleration.

Event cameras offer significant advantages over conventional frame-based counterparts, including high temporal resolution, low latency, and energy efficiency. These characteristics make them suitable for high-speed and high-dynamic range scene acquisition scenarios; however, the lack of dense intensity frames limits the direct applicability of conventional computer vision methods for scene understanding. Event-to-video (E2V) reconstruction seeks to bridge this gap by converting asynchronous event streams into a sequence of synchronous video frames. Existing E2V reconstruction methods based on convolutional neural networks and transformers operate primarily in the spatial domain and often struggle to recover fine structural details while suppressing severe reconstruction artifacts. To address these issues, we propose MSFET-E2V, a novel multiscale frequency-enhanced transformer model. At its core lies a cross-domain attention module, which fuses spatio-temporal features with frequency-aware representations derived from the discrete wavelet transform. Unlike prior methods relying solely on spatial attention, our approach effectively captures both local and global structures by taking into account low- and high-frequency components, enhancing detail preservation and robustness across various motion scenarios. Furthermore, we propose a lightweight wavelet-enhanced skip block that serves as a skip connection, facilitating artifact suppression and structural detail refinement through joint spatial-frequency domain processing. Extensive experiments demonstrate that MSFET-E2V achieves superior performance over state-of-the-art methods on multiple real-world event datasets, offering significant gains in reconstruction quality. Moreover, compared to the existing transformer-based method, our proposed model significantly reduces the number of parameters, the GPU memory usage, and inference time.