Higher Resolution, Better Generalization: Unlocking Visual Scaling in Deep Reinforcement Learning

• URL: http://arxiv.org/abs/2605.10546v1
• Date: Mon, 11 May 2026 13:23:58 GMT
• Title: Higher Resolution, Better Generalization: Unlocking Visual Scaling in Deep Reinforcement Learning
• Abstract summary: We show that observation resolution is a critical yet overlooked variable for policy learning.<n>We replace the widely used Impala encoder with global average pooling.<n>Visual scaling unlocks a 28 % performance gain for Impoola over Impala.
• Score: 2.731398851340082
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Pixel-based deep reinforcement learning agents are typically trained on heavily downsampled visual observations, a convention inherited from early benchmarks rather than grounded in principled design. In this work, we show that observation resolution is a critical yet overlooked variable for policy learning: higher-resolution inputs can substantially improve both performance and generalization, provided the network architecture can process them effectively. We find that the widely used Impala encoder, which flattens spatial features into a vector, suffers from quadratic parameter growth as resolution increases and fails to leverage the additional visual detail. Replacing this operation with global average pooling, as in the Impoola architecture, decouples parameter count from resolution and yields consistent improvements across resolutions and network widths - at their respective best conditions, visual scaling unlocks a 28 % performance gain for Impoola over Impala. These gains are strongest in environments that require precise perception of small or distant objects, and gradient saliency analysis confirms that the underlying mechanism is a more spatially localized visual attention of the policy at higher resolutions. Our results challenge the prevailing practice of aggressive input downsampling and position resolution-independent architectures as a simple, effective path toward scalable visual deep RL. To facilitate future research on resolution scaling in deep RL, we publicly release the open-source code for the Procgen-HD benchmark: https://github.com/raphajaner/procgen-hd.

Related papers

• GaussianLens: Localized High-Resolution Reconstruction via On-Demand Gaussian Densification [77.40235389999]
  We propose a generalizable network that densifies the initial 3DGS to capture fine details in a user-specified local region of interest.<n>Experiments demonstrate our method's superior performance in local fine detail reconstruction and strong scalability to images of up to $1024times1024$ resolution.
  arXiv  Detail & Related papers  (2025-09-29T23:58:49Z)
• HRSeg: High-Resolution Visual Perception and Enhancement for Reasoning Segmentation [74.1872891313184]
  HRSeg is an efficient model with high-resolution fine-grained perception.<n>It features two key innovations: High-Resolution Perception (HRP) and High-Resolution Enhancement (HRE)
  arXiv  Detail & Related papers  (2025-07-17T08:09:31Z)
• JAFAR: Jack up Any Feature at Any Resolution [53.343826346140624]
  JAFAR is a lightweight and flexible feature upsampler for Foundation Visions.<n>It enhances the spatial resolution of visual features from any Foundation Vision to an arbitrary target resolution.<n>It generalizes remarkably well to significantly higher output scales.
  arXiv  Detail & Related papers  (2025-06-10T20:53:12Z)
• HRDecoder: High-Resolution Decoder Network for Fundus Image Lesion Segmentation [12.606794661369959]
  We propose HRDecoder, a simple High-Resolution Decoder network for fundus lesion segmentation.<n>It integrates a high-resolution representation learning module to capture fine-grained local features and a high-resolution fusion module to fuse multi-scale predictions.<n>Our method effectively improves the overall segmentation accuracy of fundus lesions while consuming reasonable memory and computational overhead, and maintaining satisfying inference speed.
  arXiv  Detail & Related papers  (2024-11-06T15:13:31Z)
• PointHR: Exploring High-Resolution Architectures for 3D Point Cloud Segmentation [77.44144260601182]
  We explore high-resolution architectures for 3D point cloud segmentation. We propose a unified pipeline named PointHR, which includes a knn-based sequence operator for feature extraction and a differential resampling operator. To evaluate these architectures for dense point cloud analysis, we conduct thorough experiments using S3DIS and ScanNetV2 datasets.
  arXiv  Detail & Related papers  (2023-10-11T09:29:17Z)
• Low-Resolution Self-Attention for Semantic Segmentation [93.30597515880079]
  We introduce the Low-Resolution Self-Attention (LRSA) mechanism to capture global context at a significantly reduced computational cost.<n>Our approach involves computing self-attention in a fixed low-resolution space regardless of the input image's resolution.<n>We demonstrate the effectiveness of our LRSA approach by building the LRFormer, a vision transformer with an encoder-decoder structure.
  arXiv  Detail & Related papers  (2023-10-08T06:10:09Z)
• HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation [14.81943833870932]
  We present an improvedDepthNet, HR-Depth, with two effective strategies. Using Resnet-18 as the encoder, HR-Depth surpasses all pre-vious state-of-the-art(SoTA) methods with the least param-eters at both high and low resolution.
  arXiv  Detail & Related papers  (2020-12-14T09:15:15Z)
• PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models [77.32079593577821]
  PULSE (Photo Upsampling via Latent Space Exploration) generates high-resolution, realistic images at resolutions previously unseen in the literature. Our method outperforms state-of-the-art methods in perceptual quality at higher resolutions and scale factors than previously possible.
  arXiv  Detail & Related papers  (2020-03-08T16:44:31Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.