An official website of the United States government
We introduce a method for novel view synthesis given only a single wide-baseline stereo image pair. In this challenging regime, 3D scene points are regularly observed only once, requiring prior-based reconstruction of scene geometry and appearance. We find that existing approaches to novel view synthesis from sparse observations fail due to recovering incorrect 3D geometry and due to the high cost of differentiable rendering that precludes their scaling to large-scale training. We take a step towards resolving these shortcomings by formulating a multi-view transformer encoder, proposing an efficient, image-space epipolar line sampling scheme to assemble image features for a target ray, and a lightweight cross-attention-based renderer. Our contributions enable training of our method on a large-scale real-world dataset of indoor and outdoor scenes. We demonstrate that our method learns powerful multi-view geometry priors while reducing both rendering time and memory footprint. We conduct extensive comparisons on held-out test scenes across two real-world datasets, significantly outperforming prior work on novel view synthesis from sparse image observations and achieving multi-view-consistent novel view synthesis.
more »
« less
PROTOTYPE MEMORY AND ATTENTION MECHANISMS FOR FEW SHOT IMAGE GENERATION
Recent discoveries indicate that the neural codes in the superficial layers of the primary visual cortex (V1) of macaque monkeys are complex, diverse, and super-sparse. This leads us to ponder the computational advantages and functional role of these “grandmother cells." Here, we propose that such cells can serve as prototype memory priors that bias and shape the distributed feature processing during the image generation process in the brain. These memory prototypes are learned by momentum online clustering and are utilized through a memory-based attention operation. Integrating this mechanism, we propose Memory Concept Attention (MoCA) to improve few-shot image generation quality. We show that having a prototype memory with attention mechanisms can improve image synthesis quality, learn interpretable visual concept clusters, and improve the robustness of the model. Our results demonstrate the feasibility of the idea that these super-sparse complex feature detectors can serve as prototype memory priors for modulating the image synthesis processes in the visual system
more »
« less
- Award ID(s):
- 1816568
- PAR ID:
- 10357969
- Date Published:
- Journal Name:
- Proceedings of the Eleventh International Conference on Learning Representations
- Volume:
- 18
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
We introduce Optimal Eye Surgeon (OES), a framework for pruning and training deep image generator networks. Typically, untrained deep convolutional networks, which include image sampling operations, serve as effective image priors (Ulyanov et al., 2018). However, they tend to overfit to noise in image restoration tasks due to being overparameterized. OES addresses this by adaptively pruning networks at random initialization to a level of underparameterization. This process effectively captures low-frequency image components even without training, by just masking. When trained to fit noisy images, these pruned subnetworks, which we term Sparse-DIP, resist overfitting to noise. This benefit arises from underparameterization and the regularization effect of masking, constraining them in the manifold of image priors. We demonstrate that subnetworks pruned through OES surpass other leading pruning methods, such as the Lottery Ticket Hypothesis, which is known to be suboptimal for image recovery tasks (Wu et al., 2023). Our extensive experiments demonstrate the transferability of OES-masks and the characteristics of sparse-subnetworks for image generation.more » « less
-
Generative models learned from training using deep learning methods can be used as priors in under-determined inverse problems, including imaging from sparse set of measurements. In this paper, we present a novel hierarchical deep-generative model MrSARP for SAR imagery that can synthesize SAR images of a target at different resolutions jointly. MrSARP is trained in conjunction with a critic that scores multi resolution images jointly to decide if they are realistic images of a target at different resolutions. We show how this deep generative model can be used to retrieve the high spatial resolution image from low resolution images of the same target. The cost function of the generator is modified to improve its capability to retrieve the input parameters for a given set of resolution images. We evaluate the model's performance using three standard error metrics used for evaluating super-resolution performance on simulated data and compare it to upsampling and sparsity based image super-resolution approaches.more » « less
-
Though recent years have witnessed remarkable progress in single image super-resolution (SISR) tasks with the prosperous development of deep neural networks (DNNs), the deep learning methods are confronted with the computation and memory consumption issues in practice, especially for resource-limited platforms such as mobile devices. To overcome the challenge and facilitate the real-time deployment of SISR tasks on mobile, we combine neural architecture search with pruning search and propose an automatic search framework that derives sparse super-resolution (SR) models with high image quality while satisfying the real-time inference requirement. To decrease the search cost, we leverage the weight sharing strategy by introducing a supernet and decouple the search problem into three stages, including supernet construction, compiler-aware architecture and pruning search, and compiler-aware pruning ratio search. With the proposed framework, we are the first to achieve real-time SR inference (with only tens of milliseconds per frame) for implementing 720p resolution with competitive image quality (in terms of PSNR and SSIM) on mobile platforms (Samsung Galaxy S20).more » « less
-
High-resolution Vision-Language Models (VLMs) are widely used in multimodal tasks to enhance accuracy by preserving detailed image information. However, these models often generate an excessive number of visual tokens due to the need to encode multiple partitions of a high-resolution image input. Processing such a large number of visual tokens poses significant computational challenges, particularly for resource-constrained commodity GPUs. To address this challenge, we propose High-Resolution Early Dropping (HiRED), a plug-and-play token-dropping method designed to operate within a fixed token budget. HiRED leverages the attention of CLS token in the vision transformer (ViT) to assess the visual content of the image partitions and allocate an optimal token budget for each partition accordingly. The most informative visual tokens from each partition within the allocated budget are then selected and passed to the subsequent Large Language Model (LLM). We showed that HiRED achieves superior accuracy and performance, compared to existing token-dropping methods. Empirically, HiRED-20% (i.e., a 20% token budget) on LLaVA-Next-7B achieves a 4.7x increase in token generation throughput, reduces response latency by 78%, and saves 14% of GPU memory for single inference on an NVIDIA TESLA P40 (24 GB). For larger batch sizes (e.g., 4), HiRED-20% prevents out-of-memory errors by cutting memory usage by 30%, while preserving throughput and latency benefits.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
