cs.CVApr 3, 2026

Hierarchical Awareness Adapters with Hybrid Pyramid Feature Fusion for Dense Depth Prediction

Authors: Wuqi Su, Huilun Song, Chen Zhao, Chi Xu

Organizations: Zhejiang Gongshang University

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Monocular depth estimation from a single RGB image remains a fundamental challenge in computer vision due to inherent scale ambiguity and the absence of explicit geometric cues. Existing approaches typically rely on increasingly complex network architectures to regress depth maps, which escalates training costs and computational overhead without fully exploiting inter-pixel spatial dependencies. We propose a multilevel perceptual conditional random field (CRF) model built upon the Swin Transformer backbone that addresses these limitations through three synergistic innovations: (1) an adaptive hybrid pyramid feature fusion (HPF) strategy that captures both short-range and long-range dependencies by combining multi-scale spatial pyramid pooling with biaxial feature aggregation, enabling effective integration of global and local contextual information; (2) a hierarchical awareness adapter (HA) that enriches cross-level feature interactions within the encoder through lightweight broadcast modules with learnable dimensional scaling, reducing computational complexity while enhancing representational capacity; and (3) a fully-connected CRF decoder with dynamic scaling attention that models fine-grained pixel-level spatial relationships, incorporating a bias learning unit to prevent extreme-value collapse and ensure stable training. Extensive experiments on NYU Depth v2, KITTI, and MatterPort3D datasets demonstrate that our method achieves state-of-the-art performance, reducing Abs Rel to 0.088 (−-7.4%) and RMSE to 0.316 (−-5.4%) on NYU Depth v2, while attaining near-perfect threshold accuracy (δ<1.253≈99.8%δ< 1.25^3 \approx 99.8\%) on KITTI with only 194M parameters and 21ms inference time.

Figures & tables

MethodAbs Rel ↓\downarrowSq Rel ↓\downarrowRMSE ↓\downarrowlog RMSE ↓\downarrowδ<1.25↑\boldsymbol{\delta<1.25}\uparrowδ<1.252↑\boldsymbol{\delta<1.25^{2}}\uparrowδ<1.253↑\boldsymbol{\delta<1.25^{3}}\uparrow
VNL [ 61 ]0.108−-0.416−-0.8750.9760.994
DAV [ 24 ]0.108−-0.412−-0.8820.9800.996
DPT* [ 40 ]0.110−-0.357−-0.9040.9880.998
TransDepth [ 58 ]0.106−-0.365−-0.9000.9830.996
ASN [ 34 ]0.101−-0.377−-0.8900.982−-
PackNet-SAN* [ 20 ]0.106−-0.393−-0.8920.9790.995
Table 1: Quantitative comparison on the NYU Depth v2 dataset. Bold and underline denote best and second-best results. “ −- ” indicates unreported results. “*” indicates use of extra training data.
Methodcap/mAbs Rel ↓\downarrowSq Rel ↓\downarrowRMSE ↓\downarrowlog RMSE ↓\downarrowδ<1.25↑\boldsymbol{\delta<1.25}\uparrowδ<1.252↑\boldsymbol{\delta<1.25^{2}}\uparrowδ<1.253↑\boldsymbol{\delta<1.25^{3}}\uparrow
BTS [ 28 ]0–800.0590.2412.7560.0900.9560.9930.998
DPT* [ 40 ]0–800.062−-2.5730.0920.9590.9950.999
PWA [ 29 ]0–800.0600.2212.6040.0930.9580.9940.999
AdaBins [ 4 ]0–800.0580.1902.3600.0880.9640.9950.999
BinsFormer [ 31 ]0–800.0520.1512.0980.0790.975−-−-
NeWCRFs [ 63 ]0–800.0520.1552.1290.0790.9740.9970.999
Table 2: Quantitative results on the Eigen split of the KITTI dataset. Bold and underline denote best and second-best results.
ExpHPHAFCAbs Rel ↓\downarrowRMSE ↓\downarrowδ<1.25↑\boldsymbol{\delta<1.25}\uparrowδ<1.253↑\boldsymbol{\delta<1.25^{3}}\uparrow
A×\times×\times✓\checkmark0.0930.3280.9240.998
B×\times✓\checkmark×\times0.0900.3220.9260.998
C✓\checkmark×\times×\times0.0890.3180.9300.998
D×\times×\times✓\checkmark0.0920.3270.9240.998
E✓\checkmark×\times✓\checkmark0.0910.3250.9250.998
F✓\checkmark✓\checkmark×\times0.0880.3160.9320.998
Table 3: Ablation study on NYU Depth v2. HP: Hybrid Pyramid Feature Fusion; HA: Hierarchical Awareness Adapter; FC: Fully-Connected Decoding.
ModuleAbs Rel ↓\downarrowRMSE ↓\downarrowLog10 ↓\downarrowSq Rel ↓\downarrow
MSF only0.0930.3190.0400.042
BAF only0.0910.3210.0420.045
MSF + BAF0.0890.3180.0380.041
1MSF ( [1,2,6][1,2,6] )0.0940.3190.0400.043
2MSF ( [2,3,6][2,3,6] )0.0930.3200.0400.044
Table 4: Feature fusion module comparison.
TypeMethodParams/MTime/msAbs Rel ↓\downarrowRMSE ↓\downarrow
SFNeWCRFs270280.0520.155
BinsFormer2551230.0520.151
ZoeDepth2141770.0540.189
DDP207720.0500.148
MFMany-Depth3854880.0600.342
TC-Depth1973760.0710.330
Table 5: Performance evaluation on KITTI (352 ×\times 1216, NVIDIA RTX 3090). SF: single-frame; MF: multi-frame.
More in this topic

Monocular Depth Estimation

Monocular depth estimation infers scene geometry and distance from a single camera image, enabling 3D perception without stereo or depth sensors.

See all 135 papers

Explore similar work

CardsList
May 11, 2026

Efficient Hybrid CNN-GNN Architecture for Monocular Depth Estimation

We present GraphDepth, a monocular depth estimation architecture that synergistically integrates Graph Neural Networks (GNNs) within a convolutional encoder-decoder framework. Our approach embeds efficient GraphSAGE layers at multiple scales of a ResNet-101 U-Net backbone, enabling explicit modeling of long-range spatial relationships that lie beyond the receptive field of local convolutions. Key technical contributions include: (1) batch-parallelized graph construction with configurable k-NN and grid-based adjacency for scalable training; (2) multi-scale GraphSAGE integration at bottleneck and decoder stages (1/32, 1/16, 1/8 resolution) to propagate global context throughout the feature hierarchy; (3) channel-attention gated skip connections that adaptively weight encoder features before fusion; and (4) heteroscedastic uncertainty estimation via a dedicated aleatoric uncertainty head, enabling confidence-aware loss weighting during optimization. Unlike transformer-based hybrids, which suffer from quadratic complexity in sequence length, GraphDepth scales linearly with spatial resolution while achieving comparable global receptive fields through iterative message passing. Experiments on NYU Depth V2, WHU Aerial, ETH3D, and Mid-Air benchmarks demonstrate competitive accuracy within 4.6% of state-of-the-art transformers on indoor scenes with substantially lower computational cost (25 FPS vs 9 FPS, 3.8 GB vs 8.8 GB VRAM). GraphDepth achieves the best reported result on WHU Aerial (RMSE 8.24 m) and exhibits superior zero-shot cross-domain transfer to the Mid-Air synthetic aerial dataset, validating the generalization power of explicit relational reasoning for depth estimation.

Ishan Narayan

IMCS Lab, CSIR-CSIO · Chandigarh 160030, India

Graph Neural NetworksDepth EstimationMonocular Depth Estimation
AskPDF
Sep 21, 2026

CMAMBADEPTH: Self-supervised Monocular Depth Estimation with Channel Mamba and Hybrid Attention

Accurate monocular depth estimation serves as a core enabler for single camera scene understanding. However, existing self-supervised monocular depth estimation methods generally suffer from the bottleneck of inefficient cross-scale information interaction and difficulty in balancing local and global spatial modeling. In this paper, we propose CMambaDepth, a self-supervised framework that achieves efficient multi-scale feature fusion and fine-grained contextual modeling via channel-wise selective state propagation. Specifically, Bidirectional Channel Mamba (Bi-CMamba) aligns encoder features across scales and enables bidirectional information exchange among ordered scale groups. Unidirectional Channel Mamba (Uni-CMamba) progressively aggregates decoder features and retains fine-grained scale groups through a group selection mechanism for subsequent fusion. Furthermore, a Hybrid Attention Module (HAM) is introduced to combine large-kernel local context and Manhattan self-attention for complementary spatial modeling. Experimental results demonstrate that our method achieves highly competitive performance. Specifically, our model achieves an AbsRel of 0.094 and an RMSE of 4.156 on KITTI, and an AbsRel of 0.140 on DDAD. In the zero-shot cross-dataset generalization test on NYUv2, it attains an AbsRel of 0.232, outperforming the baseline RA-Depth by 7.2%.

Xuezhi Xiang, Jiayao Liu, Heqi Xiang +3

School of Information and Communication Engineering, Harbin Engineering University, Harbin 150001, China · Key Laboratory of Advanced Marine Communication and Information Technology, Harbin 150001, China · Department of Computer Science, University of Toronto, Toronto ON M5S 2E4, Canada +1

Monocular Depth EstimationMulti-Scale Feature FusionVisual SSMsSelf-Supervised Monocular Depth Estimation
AskPDF
Aug 17, 2026

PXDepth: Pixel-Space Modeling for Structure Preserving Monocular Depth Estimation

Recent monocular depth estimators achieve strong zero-shot generalization, yet often struggle to preserve fine-grained structures and object boundaries. We attribute this limitation to the prevalent combination of large-patch ViT encoders and convolutional decoders, as coarse tokenization can weaken pixel-level cues that upsampling cannot fully recover. To address this issue, we propose PXDepth, a discriminative monocular depth model that separates global context modeling from pixel-level depth prediction. Specifically, a large-patch ViT captures global scene context, while a pixel-space predictor composed of Context-Modulated Pixel Transformer blocks maintains high-resolution spatial representations throughout depth estimation. This design preserves fine structures and sharp boundaries without sacrificing global depth consistency. Across diverse zero-shot benchmarks, PXDepth combines faithful local geometry with competitive global depth accuracy while remaining efficient at inference. Our code and model are available at https://yuanzhy29.github.io/PXDepth-Page/.

Zhiyuan Yuan, Guanying Chen, Lingteng Qiu +3

1Sun Yat-sen University · 3CUHKSZ · 2Shenzhen-FNii

Depth EstimationVision TransformerMonocular Depth Estimation
AskPDF