AECR-Net: Compact Dehazing Deep Network

• AECR-Net is a compact, end-to-end autoencoder deep network for dehazing that integrates attention, contrastive regularization, adaptive mixup, and dynamic feature enhancement.
• It leverages a feature-attention backbone with U-Net skip connections and dynamic feature enhancement, achieving state-of-the-art PSNR and SSIM on both synthetic and real-world benchmarks.
• The model’s efficient design, with only 2.6–3M parameters, enables robust dehazing for adverse visibility conditions and potential applications in other image restoration tasks.

AECR-Net is a compact, end-to-end autoencoder-style deep network designed for single image dehazing and enhancement under challenging visibility conditions such as haze and smoke. The model integrates a feature-attention backbone, contrastive regularization, adaptive mixup, and a dynamic feature enhancement mechanism. AECR-Net has demonstrated state-of-the-art quantitative and qualitative results on both synthetic and real-world haze and smoke benchmarks, with practical utility further validated for gauge image interpretation in adverse visibility applications (Ramírez-Agudelo et al., 15 Jan 2026, Wu et al., 2021).

1. Architectural Components

At its core, AECR-Net employs an autoencoder architecture augmented with various feature fusion and regularization modules:

• Encoder Stem: Processes an RGB input $I_{in} \in \mathbb{R}^{3 \times H \times W}$ through a $7 \times 7$ convolution (64 channels), ReLU, and InstanceNorm, followed by downsampling (stride-2, 128 channels).
• Bottleneck: Comprises $N$ stacked AECR-Blocks, each integrating FFA-Net's Feature Attention (FA) block—composed of Channel-Attention (CA) and Pixel-Attention (PA) sub-modules—with local residuals: $X_{out} = X_{in} + FA(X_{in})$. Every $M$ blocks (typically $M = 2$), attributes a Dynamic Feature Enhancement (DFE) operation to adaptively re-scale channel statistics.
• Decoder: Mirrors the encoder with upsampling (using nearest-neighbor or transposed convolutions), ReLU, InstanceNorm, and final $3 \times 3$ convolution projecting to $3$ output channels with sigmoid output activation.
• Skip Connections: Employs U-Net-style skip connections to link encoder and decoder layers at corresponding spatial resolutions, enhancing feature fusion.
• Adaptive Mixup Module: During training only, generates synthetic feature-space interpolations (mix-of-pairs) between clean and hazy encodings to augment representation diversity for contrastive learning.

The model contains approximately $2.6$–$3$ million parameters, making it compact relative to other state-of-the-art dehazing networks (Wu et al., 2021).

2. Mathematical Formulation and Losses

AECR-Net’s objective combines image reconstruction fidelity with a pixel-level contrastive constraint:

• Reconstruction Loss: Per-pixel $7 \times 7$0 distance between dehazed output $7 \times 7$1 and ground truth $7 \times 7$2:

$7 \times 7$3

• Contrastive Loss (InfoNCE): Given representations $7 \times 7$4 extracted via a pre-trained feature extractor (e.g., VGG-19), AECR-Net enforces:

$7 \times 7$5

where $7 \times 7$6 denotes cosine similarity and $7 \times 7$7 is a temperature hyperparameter.

• Total Objective: Weighted combination,

$7 \times 7$8

with typical weighting $7 \times 7$9 or, in some variants, $N$0 (Wu et al., 2021).

Contrastive regularization constrains network outputs to lie closer to ground truth features while diverging from hazy inputs, narrowing the feasible restoration manifold (Wu et al., 2021).

3. Feature Engineering: Attention, Mixup, DFE

• Feature Attention (FA): Inherits both channel and pixel attention from FFA-Net to emphasize discriminative statistics at multiple representation levels.
• Adaptive Mixup: Parameterizes skip-level feature fusion via two learned interpolations:

$N$1

blending encoder and decoder activations to preserve spatial detail and yield sharper reconstructions.

• Dynamic Feature Enhancement (DFE): Utilizes two stacked modulated deformable convolutions:

$N$2

where $N$3 are learned offsets, $N$4 are modulation masks, and $N$5 are kernel weights, adaptively expanding the receptive field.

Ablation experiments demonstrate each module’s contribution: DFE alone yields $N$6$N$7 dB PSNR, adaptive mixup provides $N$8$N$9 dB, and full contrastive regularizer delivers %%%%3$I_{in} \in \mathbb{R}^{3 \times H \times W}$3%%%%1 dB over positives-only (Wu et al., 2021).

4. Training Protocols and Evaluation

Datasets:

• For gauge image enhancement, synthetic data is generated in Unreal Engine 5.1.1 with realistic global illumination, custom 3D gauge meshes, and both Exponential Height Fog and GPU-simulated smoke particles. Each scene yields images across 10 haze and 10 smoke density levels, with one clear reference.
• RESIDE benchmark is used for generic single-image dehazing, with SOTS (indoor) test set for evaluation, and Dense-Haze and NH-HAZE for real-world validation.

Hyperparameters:

• Optimizer: Adam $X_{out} = X_{in} + FA(X_{in})$2
• Initial LR: $X_{out} = X_{in} + FA(X_{in})$3 or $X_{out} = X_{in} + FA(X_{in})$4, stepped or cosine annealing
• Batch size: $X_{out} = X_{in} + FA(X_{in})$5 (gauge) or $X_{out} = X_{in} + FA(X_{in})$6 (RESIDE)
• Epochs: $X_{out} = X_{in} + FA(X_{in})$7
• Loss weights: $X_{out} = X_{in} + FA(X_{in})$8, $X_{out} = X_{in} + FA(X_{in})$9
• Feature extraction for contrastive loss layers: VGG-19, layers indexed {1,3,5,9,13} with progressively increasing weights

No data augmentations are applied beyond density-level variation.

5. Quantitative Performance

Dataset	Method	PSNR (dB)	SSIM	Parameters
Haze (Gauge)	AECR-Net	∼44	0.98	3M
	FFA-Net	∼30	0.96	-
	BCCR	∼12	0.65	-
Smoke (Gauge)	AECR-Net	∼37	0.96	3M
	FFA-Net	∼26	0.94	-
	BCCR	∼9	0.55	-
RESIDE/SOTS	AECR-Net	37.17	0.990	2.6M
	FFA-Net	36.39	0.989	4.68M
	(Others)	∼30	∼0.97	>3M

On both synthetic and real-world datasets, AECR-Net matches or outperforms prior methods in PSNR and SSIM, despite its low model size. In custom smoke and haze gauge datasets, AECR-Net improves PSNR by roughly $M$0 dB over FFA-Net (Ramírez-Agudelo et al., 15 Jan 2026, Wu et al., 2021).

6. Component-Level Analysis and Practical Impact

• Contrastive Regularization: Adding $M$1 increases average PSNR by about $M$2 dB (gauge enhancement) and $M$3 dB (RESIDE SOTS).
• Dynamic Feature Enhancement: Disabling this module leads to $M$40.01 drop in SSIM and $M$51.7 dB PSNR loss.
• Adaptive Mixup: Reduces overfitting to mid-level densities, improving generalization to unseen visibility conditions.
• FA-Block Depth: Increasing the number of FA blocks to $M$6 improves performance, with diminishing returns beyond $M$7.
• Parameter Efficiency: AECR-Net’s compactness (2.6–3M parameters) provides substantial computational and memory advantages relative to previous deep dehazing models (Wu et al., 2021).

In infrastructure and emergency response settings, AECR-Net’s enhanced outputs enable more robust post-processing pipelines for automatic and autonomous gauge interpretation, critical in haze- and smoke-obscured environments (Ramírez-Agudelo et al., 15 Jan 2026).

7. Availability and Extensions

The official implementation of AECR-Net is available at https://github.com/GlassyWu/AECR-Net. Contrastive regularization has shown universality: adding CR to alternative single-image dehazing architectures yields consistent PSNR and SSIM gains without increased inference costs, supporting AECR-Net’s utility as a general model design motif.

A plausible implication is that AECR-Net’s architecture and learning protocol may generalize to deblurring, low-light enhancement, and other single-image restoration tasks where information degradation results from complex, content-dependent imaging perturbations.

---

Key References:

• "Enhancing the quality of gauge images captured in smoke and haze scenes through deep learning" (Ramírez-Agudelo et al., 15 Jan 2026)
• "Contrastive Learning for Compact Single Image Dehazing" (Wu et al., 2021)