cs.CVNov 11, 2026

FoveaDet: Software-Hardware Co-Design of Foveated CNN-Transformer for Object Detection

Authors: Deepak Mewada, Apporva Saraf, Ashwin Lele

Organizations: Centre for Machine Intelligence and Data Science (CMInDS), IIT Bombay, India · Department of Electrical Engineering, IIT Delhi, India

Abstract

Deploying real-time object detection on power-constrained edge platforms is challenging as processing full-resolution frames is often computationally redundant. Inspired by biological foveation which assigns variable resolution based on visual saliency, we propose a novel software-hardware co-design framework called FoveaDet. On the software side, FoveaDet dynamically identifies salient regions to predict fixation points, followed by aggressive patch pruning in Vision Transformers (ViT) to discard irrelevant background data. To mitigate slowdowns from sequential patch generation, we implement optimized interpolation and patch selection algorithms. Evaluated on standard GPUs, this foveated approach reduces floating-point operations by 65.4% to 92.2% and accelerates inference by 4.76x, while suffering a negligible mAP drop compared to dense baselines. To eliminate hardware bottlenecks, we complement the algorithm with a custom RISC-V core featuring specialized instructions to accelerate the foveated processing pipeline. FoveaDet provides a scalable, efficient architecture capable of unlocking advanced vision processing on severely constrained mobile platforms.

Figures & tables

Explore similar work

Oct 29, 2025

RT-DETRv4: Painlessly Furthering Real-Time Object Detection with Vision Foundation Models

Real-time object detection has achieved substantial progress through meticulously designed architectures and optimization strategies. However, the pursuit of high-speed inference via lightweight network designs often leads to degraded feature representation, which hinders further performance improvements and practical on-device deployment. In this paper, we propose a cost-effective and highly adaptable distillation framework that harnesses the rapidly evolving capabilities of Vision Foundation Models (VFMs) to enhance lightweight object detectors. Given the significant architectural and learning objective disparities between VFMs and resource-constrained detectors, achieving stable and task-aligned semantic transfer is challenging. To address this, on one hand, we introduce a Deep Semantic Injector (DSI) module that facilitates the integration of high-level representations from VFMs into the deep layers of the detector. On the other hand, we devise a Gradient-guided Adaptive Modulation (GAM) strategy, which dynamically adjusts the intensity of semantic transfer based on gradient norm ratios. Without increasing deployment and inference overhead, our approach painlessly delivers striking and consistent performance gains across diverse DETR-based models, underscoring its practical utility for real-time detection. Our new model family, RT-DETRv4, achieves state-of-the-art results on COCO, attaining AP scores of 49.8/53.7/55.4/57.049.8/53.7/55.4/57.0 at corresponding speeds of 273/169/124/78273/169/124/78 FPS. Code is publicly available at https://github.com/RT-DETRs/RT-DETRv4.
May 11, 2026

MicroViTv2: Beyond the FLOPS for Edge Energy-Friendly Vision Transformers

The Vision Transformer (ViT) achieves remarkable accuracy across visual tasks but remains computationally expensive for edge deployment. This paper presents MicroViTv2, a lightweight Vision Transformer optimized for real-device efficiency. Built upon the original MicroViT, the proposed model is designed based on reparameterized design, specifically Reparameterized Patch Embedding (RepEmbed) and Reparameterized Depth-Wise convolution mixer (RepDW) for faster inference, and introduces the Single Depth-Wise Transposed Attention (SDTA) to capture long-range dependencies with minimal redundancy. Despite slightly higher FLOPs, MicroViTv2 improves accuracy up to 0.5% compared to its predecessor and surpassing MobileViTv2, EdgeNeXt, and EfficientViT while maintaining fast inference and high energy efficiency on Jetson AGX Orin. Experiments on ImageNet-1K and COCO demonstrate that hardware-aware design and structural re-parameterization are key to achieving high accuracy and low energy consumption, validating the need to evaluate efficiency beyond FLOPs. Code is available at https://github.com/novendrastywn/MicroViT.
Jun 30, 2026

FlexViT: A Flexible FPGA-based Accelerator for Edge Vision Transformers

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