A Generative Adversarial Network (GAN) with generator G trained to model the prior of images has been shown to perform better than sparsity-based regularizers in illposed inverse problems. Here, we propose a new method of deploying a GAN-based prior to solve linear inverse problems using projected gradient descent (PGD). Our method learns a network-based projector for use in the PGD algorithm, eliminating expensive computation of the Jacobian of G. Experiments show that our approach provides a speed-up of 60-80× over earlier GAN-based recovery methods along with better accuracy. Our main theoretical result is that if the measurement matrix is moderately conditioned on the manifold range(G) and the projector is δ-approximate, then the algorithm is guaranteed to reach O(δ) reconstruction error in O(log(1/δ)) steps in the low noise regime. Additionally, we propose a fast method to design such measurement matrices for a given G. Extensive experiments demonstrate the efficacy of this method by requiring 5-10× fewer measurements than random Gaussian measurement matrices for comparable recovery performance. Because the learning of the GAN and projector is decoupled from the measurement operator, our GAN-based projector and recovery algorithm are applicable without retraining to all linear inverse problems, as confirmed by experiments on compressed sensing, super-resolution, and inpainting.

Temporal modeling in videos is a fundamental yet challenging problem in computer vision. In this paper, we propose a novel Temporal Bilinear (TB) model to capture the temporal pairwise feature interactions between adjacent frames. Compared with some existing temporal methods which are limited in linear transformations, our TB model considers explicit quadratic bilinear transformations in the temporal domain for motion evolution and sequential relation modeling. We further leverage the factorized bilinear model in linear complexity and a bottleneck network design to build our TB blocks, which also constrains the parameters and computation cost. We consider two schemes in terms of the incorporation of TB blocks and the original 2D spatial convolutions, namely wide and deep Temporal Bilinear Networks (TBN). Finally, we perform experiments on several widely adopted datasets including Kinetics, UCF101 and HMDB51. The effectiveness of our TBNs is validated by comprehensive ablation analyses and comparisons with various state-of-the-art methods.

No abstract

Compressive imaging systems with spatial-temporal encoding can be used to capture and reconstruct fast-moving objects. The imaging quality highly depends on the choice of encoding masks and reconstruction methods. In this paper, we present a new network architecture to jointly design the encoding masks and the reconstruction method for compressive high-frame-rate imaging. Unlike previous works, the proposed method takes full advantage of a denoising prior to provide a promising frame reconstruction. The network is also flexible enough to optimize full-resolution masks and efficient at reconstructing frames. To this end, we develop a new dense network architecture that embeds Anderson acceleration, known from numerical optimization, directly into the neural network architecture. Our experiments show the optimized masks and the dense accelerated network respectively achieve 1.5 dB and 1 dB improvements in PSNR without adding training parameters. The proposed method outperforms other state-of-the-art methods both in simulations and on real hardware. In addition, we set up a coded two-bucket camera for compressive high-frame-rate imaging, which is robust to imaging noise and provides promising results when recovering nearly 1,000 frames per second. Index Terms-high-frame-rate imaging, deep neural network, computational camera ! • This paper is under review for ICCP 2020 and the PAMI special issue on computational photography. Do not distribute. [6]. However, the encoding and decoding components of 35 the imaging system are highly interdependent. Based on 36 this observation, we focus on the joint end-to-end design of 37 encoding masks and reconstruction methods for improving 38 both encoding efficiency and reconstruction accuracy. We 39 put forward a compact end-to-end neural network that can 40 handle the mask optimization for the whole image with 41 fewer training parameters. We also show that this network 49 • We present the first work to jointly design full-reso-50 lution coding masks and reconstruction methods for 51 compressive high-frame-rate imaging using an end-52 to-end network. Our approach outperforms state-of-53 art methods by 2.2dB in PSNR. 54 • We show that the acceleration of the gradient de-55 scent algorithm is equivalent to adding dense skip 56 connections to iterative optimization-unrolling neu-57 ral networks. This speeds up training convergence 58 and helps to design a compact and efficient network 59 architecture. 60 • Experiments on both simulation and real hardware 61 demonstrate the effectiveness of our reconstruction 62 method and the designed masks. The two-bucket 63 design of our camera shows improved noise sup-64 pression and can provide promising results in re-65 constructing video of frame rates up to almost 1,000 66 frames per second. 67 2 RELATED WORK 68 Many approaches have been developed to solve the ill-69 conditioned inverse problem in CS. The existing methods 70 can be divided into model-based optimization methods, 71 deep discriminative learning methods, and unrolled itera-72 t...

achieves state-of-the-art results for HS-D imaging and that the optimized DOE outperforms alternative optical designs.