ProgResViT is an input‑adaptive Vision Transformer that processes images progressively across multiple rounds, starting with a low‑resolution image and a narrow subnetwork and refining the prediction with higher resolution and a wider subnetwork if needed. The method introduces Progress‑Conditioned Soft Gating (PSG) to share a single backbone across rounds while conditioning token fusion and layer outputs on the current round, block, and input resolution. Experiments on DeiT show improved accuracy‑compute trade‑offs compared to adaptive‑width, adaptive‑depth, and dynamic‑token baselines, and the design also benefits self‑supervised DINO representations and downstream semantic segmentation.
arXiv:2607. 02612v1 Announce Type: cross Abstract: Vision Transformers achieve strong image classification accuracy but process all image regions with nearly the same computation, even when many regions are redundant or uninformative.
arXiv:2607. 00774v1 Announce Type: cross Abstract: Recent recursive Transformer studies have primarily reused shared parameters across computation steps to construct compact, parameter-efficient models.
The paper introduces reViT, a recurrent vision transformer that uses a single Transformer block repeatedly to achieve the accuracy of a full-depth encoder while keeping inference FLOPs comparable. It achieves depth‑specific transformations by expressing each block’s feed‑forward network as a convex combination of a small shared expert bank, with a continuous depth coordinate guiding the mixture. Experiments on ImageNet‑1k and DINOv2 distillation show that reViT matches or surpasses state‑of‑the‑art models, reduces stored parameters by about 70%, and supports elastic‑depth deployment from a single checkpoint.
The paper introduces Gated Token Recurrence (GTR), a softmax‑free recurrent vision backbone that replaces global softmax attention with gated linear attention, alternating scan directions, and enhanced SwiGLU blocks. GTR is distilled from a DINOv3 teacher using only final‑layer patch‑token alignment, and achieves strong performance on COCO object detection (58.9 box AP) with very low latency (1.908 ms on an RTX 4090). The backbone also transfers to multiple dense prediction tasks and runs efficiently on edge hardware via a specialized CUDA operator and TensorRT deployment.
arXiv:2607. 06982v1 Announce Type: cross Abstract: Convolutional neural networks (CNNs) have demonstrated encouraging results in image classification tasks.
The paper introduces Recursive Block-Diagonal Coupling (RBDC), a training protocol that builds wide vision models by recursively coupling narrower, independently trained models in a parameter‑free block‑diagonal manner. RBDC allows flexible allocation of training budgets across all models and, when applied to vision transformers (DeiT) and convolutional networks (ResNet) on ImageNet, achieves a 30% reduction in FLOPs while maintaining similar test accuracies. Additionally, models trained with RBDC outperform those from existing growth methods at the same training FLOPs and serve as stronger backbones for downstream tasks such as object detection and instance segmentation.
The paper studies Graph-Guided Token Merging (G2TM), a module that reduces token count in Vision Transformers. It evaluates G2TM across multiple segmentation frameworks and decoder types, finding that its performance gains are tied to the encoder rather than the decoder. The authors report consistent reductions in GFLOPs (22‑47%) and throughput improvements (up to 74%) on ADE20K, with optimal hyperparameters depending mainly on backbone pre‑training and target dataset.
Convolutional neural networks (CNNs) have demonstrated encouraging results in image classification tasks. However, the prohibitive computational cost of CNNs hinders the deployment of CNNs onto resource-constrained embedded devices.
arXiv:2609.37631v1 Announce Type: new Abstract: Transformers are typically trained from random initialization, requiring all their capabilities to emerge from large-scale optimization. Recent work sh...
arXiv:2410.13925v2 Announce Type: replace Abstract: \textit{Nature is infinitely resolution-free}. In the context of this reality, existing diffusion models, such as Diffusion Transformers, often fac...
GramLoop is a training‑free framework that enhances frozen DINOv3 dense‑prediction models under distribution shift by adding inference computation within the visual backbone. It replays a short transformer window and uses final‑layer cosine‑Gram consistency to control each replay, propagating proposals through the frozen suffix and accepting them via a patchwise gate. Across object detection and semantic segmentation tasks, GramLoop improves performance on all five shifted benchmarks, notably raising COCO‑O mAP by +0.252 and Effective Robustness by +0.250 while maintaining clean ADE20K accuracy.
arXiv:2607.09086v2 Announce Type: replace Abstract: We present Subtoken Vision Transformer (SubViT), a selective image tokenization method for fine-grained visual recognition. Standard Vision Transfo...