Neural Architecture Search for Efficient Uncalibrated Deep Photometric Stereo

We present an automated machine learning approach for uncalibrated photometric stereo (PS). Our work aims at discovering lightweight and computationally efficient PS neural networks with excellent surface normal accuracy. Unlike previous uncalibrated deep PS networks, which are handcrafted and carefully tuned, we leverage differentiable neural architecture search (NAS) strategy to find uncalibrated PS architecture automatically. We begin by defining a discrete search space for a light calibration network and a normal estimation network, respectively. We then perform a continuous relaxation of this search space and present a gradient-based optimization strategy to find an efficient light calibration and normal estimation network. Directly applying the NAS methodology to uncalibrated PS is not straightforward as certain task-specific constraints must be satisfied, which we impose explicitly. Moreover, we search for and train the two networks separately to account for the Generalized Bas-Relief (GBR) ambiguity. Extensive experiments on the DiLiGenT dataset show that the automatically searched neural architectures performance compares favorably with the state-of-the-art uncalibrated PS methods while having a lower memory footprint.

[1]  R. Cipolla,et al.  PX-NET: Simple and Efficient Pixel-Wise Training of Photometric Stereo Networks , 2020, 2021 IEEE/CVF International Conference on Computer Vision (ICCV).

[2]  Cristian Sminchisescu,et al.  A Real-Time Online Learning Framework for Joint 3D Reconstruction and Semantic Segmentation of Indoor Scenes , 2021, IEEE Robotics and Automation Letters.

[3]  Yongtian Wang,et al.  Robust Photometric Stereo via Low-Rank Matrix Completion and Recovery , 2010, ACCV.

[4]  Takanori Maehara,et al.  Neural Inverse Rendering for General Reflectance Photometric Stereo , 2018, ICML.

[5]  Luc Van Gool,et al.  Neural Architecture Search of SPD Manifold Networks , 2020, IJCAI.

[6]  Mike J. Chantler,et al.  Can two specular pixels calibrate photometric stereo? , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.

[7]  Yasuyuki Matsushita,et al.  What Is Learned in Deep Uncalibrated Photometric Stereo? , 2020, ECCV.

[8]  Jean Ponce,et al.  Accurate, Dense, and Robust Multiview Stereopsis , 2010, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[9]  Shree K. Nayar,et al.  Diffuse structured light , 2012, 2012 IEEE International Conference on Computational Photography (ICCP).

[10]  Zhe Wu,et al.  Calibrating Photometric Stereo by Holistic Reflectance Symmetry Analysis , 2013, 2013 IEEE Conference on Computer Vision and Pattern Recognition.

[11]  Yasuyuki Matsushita,et al.  Self-Calibrating Deep Photometric Stereo Networks , 2019, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[12]  Hongdong Li,et al.  Monocular Dense 3D Reconstruction of a Complex Dynamic Scene from Two Perspective Frames , 2017, 2017 IEEE International Conference on Computer Vision (ICCV).

[13]  Edward H. Adelson,et al.  Shape estimation in natural illumination , 2011, CVPR 2011.

[14]  David J. Kriegman,et al.  Reflections on the generalized bas-relief ambiguity , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05).

[15]  Yoichi Sato,et al.  Uncalibrated photometric stereo based on elevation angle recovery from BRDF symmetry of isotropic materials , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[16]  Ashutosh Saxena,et al.  Make3D: Learning 3D Scene Structure from a Single Still Image , 2009, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[17]  Quoc V. Le,et al.  EfficientDet: Scalable and Efficient Object Detection , 2019, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[18]  Yuandong Tian,et al.  FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search , 2018, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[19]  Sergey Ioffe,et al.  Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift , 2015, ICML.

[20]  Kiriakos N. Kutulakos,et al.  A Theory of Refractive and Specular 3D Shape by Light-Path Triangulation , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.

[21]  Alok Aggarwal,et al.  Regularized Evolution for Image Classifier Architecture Search , 2018, AAAI.

[22]  Anoop Cherian,et al.  Scalable Dense Non-rigid Structure-from-Motion: A Grassmannian Perspective , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[23]  Xiangxiang Chu,et al.  Fair DARTS: Eliminating Unfair Advantages in Differentiable Architecture Search , 2019, ECCV.

[24]  Charlie C. L. Wang,et al.  Photometric stereo with near point lighting: A solution by mesh deformation , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[25]  David J. Kriegman,et al.  The Bas-Relief Ambiguity , 2004, International Journal of Computer Vision.

[26]  Suryansh Kumar,et al.  Non-Rigid Structure from Motion: Prior-Free Factorization Method Revisited , 2019, 2020 IEEE Winter Conference on Applications of Computer Vision (WACV).

[27]  Li Fei-Fei,et al.  Progressive Neural Architecture Search , 2017, ECCV.

[28]  Kun Li,et al.  GPS-Net: Graph-based Photometric Stereo Network , 2020, NeurIPS.

[29]  Quoc V. Le,et al.  Neural Architecture Search with Reinforcement Learning , 2016, ICLR.

[30]  David J. Kriegman,et al.  Photometric stereo with non-parametric and spatially-varying reflectance , 2008, 2008 IEEE Conference on Computer Vision and Pattern Recognition.

[31]  Yasuyuki Matsushita,et al.  Deep Near-Light Photometric Stereo for Spatially Varying Reflectances , 2020, ECCV.

[32]  Yasuyuki Matsushita,et al.  Deep Photometric Stereo for Non-Lambertian Surfaces , 2020, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[33]  Luc Van Gool,et al.  Uncalibrated Neural Inverse Rendering for Photometric Stereo of General Surfaces , 2020, 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[34]  Yoichi Sato,et al.  SymPS: BRDF Symmetry Guided Photometric Stereo for Shape and Light Source Estimation , 2018, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[35]  Jian Sun,et al.  Identity Mappings in Deep Residual Networks , 2016, ECCV.

[36]  Kirthevasan Kandasamy,et al.  Neural Architecture Search with Bayesian Optimisation and Optimal Transport , 2018, NeurIPS.

[37]  Paolo Favaro,et al.  A Closed-Form, Consistent and Robust Solution to Uncalibrated Photometric Stereo Via Local Diffuse Reflectance Maxima , 2013, International Journal of Computer Vision.

[38]  Jiaya Jia,et al.  Efficient photometric stereo on glossy surfaces with wide specular lobes , 2008, 2008 IEEE Conference on Computer Vision and Pattern Recognition.

[39]  Steven M. Seitz,et al.  Shape and spatially-varying BRDFs from photometric stereo , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.

[40]  Ping Tan,et al.  A Benchmark Dataset and Evaluation for Non-Lambertian and Uncalibrated Photometric Stereo , 2019, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[41]  In-So Kweon,et al.  Partial Sum Minimization of Singular Values in RPCA for Low-Level Vision , 2013, 2013 IEEE International Conference on Computer Vision.

[42]  Luc Van Gool,et al.  Trilevel Neural Architecture Search for Efficient Single Image Super-Resolution , 2021, ArXiv.

[43]  Ufuk Sakarya,et al.  Three-dimensional surface reconstruction for cartridge cases using photometric stereo. , 2008, Forensic science international.

[44]  Hang Xu,et al.  Auto-FPN: Automatic Network Architecture Adaptation for Object Detection Beyond Classification , 2019, 2019 IEEE/CVF International Conference on Computer Vision (ICCV).

[45]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[46]  Kai Han,et al.  PS-FCN: A Flexible Learning Framework for Photometric Stereo , 2018, ECCV.

[47]  Andrew Zisserman,et al.  SilNet : Single- and Multi-View Reconstruction by Learning from Silhouettes , 2017, BMVC.

[48]  Athinodoros S. Georghiades,et al.  Incorporating the Torrance and Sparrow model of reflectance in uncalibrated photometric stereo , 2003, Proceedings Ninth IEEE International Conference on Computer Vision.

[49]  Tianqi Chen,et al.  Empirical Evaluation of Rectified Activations in Convolutional Network , 2015, ArXiv.

[50]  Zhe Wu,et al.  A Benchmark Dataset and Evaluation for Non-Lambertian and Uncalibrated Photometric Stereo , 2019, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[51]  David J. Kriegman,et al.  Isotropy, Reciprocity and the Generalized Bas-Relief Ambiguity , 2007, 2007 IEEE Conference on Computer Vision and Pattern Recognition.

[52]  Luca Antiga,et al.  Automatic differentiation in PyTorch , 2017 .

[53]  Katsushi Ikeuchi,et al.  Ieee Transactions on Pattern Analysis and Machine Intelligence Bi-polynomial Modeling of Low-frequency Reflectances , 2022 .

[54]  Jiansheng Peng,et al.  Single image 3D object reconstruction based on deep learning: A review , 2020, Multimedia Tools and Applications.

[55]  Hongdong Li,et al.  Superpixel Soup: Monocular Dense 3D Reconstruction of a Complex Dynamic Scene , 2019, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[56]  King-Sun Fu,et al.  IEEE Transactions on Pattern Analysis and Machine Intelligence Publication Information , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[57]  Kiyoharu Aizawa,et al.  Photometric Stereo Using Sparse Bayesian Regression for General Diffuse Surfaces , 2014, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[58]  Yiming Yang,et al.  DARTS: Differentiable Architecture Search , 2018, ICLR.

[59]  David J. Kriegman,et al.  Resolving the Generalized Bas-Relief Ambiguity by Entropy Minimization , 2007, 2007 IEEE Conference on Computer Vision and Pattern Recognition.

[60]  Vijay Vasudevan,et al.  Learning Transferable Architectures for Scalable Image Recognition , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[61]  Satoshi Ikehata,et al.  CNN-PS: CNN-based Photometric Stereo for General Non-Convex Surfaces , 2018, ECCV.

[62]  Kiyoharu Aizawa,et al.  Photometric Stereo Using Constrained Bivariate Regression for General Isotropic Surfaces , 2014, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[63]  Andrew W. Fitzgibbon,et al.  KinectFusion: Real-time dense surface mapping and tracking , 2011, 2011 10th IEEE International Symposium on Mixed and Augmented Reality.

[64]  Robert J. Woodham,et al.  Photometric method for determining surface orientation from multiple images , 1980 .

[65]  Yasuyuki Matsushita,et al.  Self-calibrating photometric stereo , 2010, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[66]  Ignas Budvytis,et al.  A CNN Based Approach for the Near-Field Photometric Stereo Problem , 2020, BMVC.

[67]  Hongdong Li,et al.  Dense Depth Estimation of a Complex Dynamic Scene without Explicit 3D Motion Estimation , 2019, 1902.03791.

[68]  Quoc V. Le,et al.  Large-Scale Evolution of Image Classifiers , 2017, ICML.

[69]  Li Fei-Fei,et al.  Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation , 2019, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[70]  Suryansh Kumar,et al.  Jumping Manifolds: Geometry Aware Dense Non-Rigid Structure From Motion , 2019, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[71]  Quoc V. Le,et al.  Efficient Neural Architecture Search via Parameter Sharing , 2018, ICML.