Papers
Topics
Authors
Recent
Search
2000 character limit reached

HOSC: A Periodic Activation Function for Preserving Sharp Features in Implicit Neural Representations

Published 20 Jan 2024 in cs.NE, cs.CV, cs.GR, and cs.LG | (2401.10967v1)

Abstract: Recently proposed methods for implicitly representing signals such as images, scenes, or geometries using coordinate-based neural network architectures often do not leverage the choice of activation functions, or do so only to a limited extent. In this paper, we introduce the Hyperbolic Oscillation function (HOSC), a novel activation function with a controllable sharpness parameter. Unlike any previous activations, HOSC has been specifically designed to better capture sudden changes in the input signal, and hence sharp or acute features of the underlying data, as well as smooth low-frequency transitions. Due to its simplicity and modularity, HOSC offers a plug-and-play functionality that can be easily incorporated into any existing method employing a neural network as a way of implicitly representing a signal. We benchmark HOSC against other popular activations in an array of general tasks, empirically showing an improvement in the quality of obtained representations, provide the mathematical motivation behind the efficacy of HOSC, and discuss its limitations.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (37)
  1. A survey on modern trainable activation functions. Neural networks : the official journal of the International Neural Network Society, 138:14–32, 2020. URL https://api.semanticscholar.org/CorpusID:218487292.
  2. Sal: Sign agnostic learning of shapes from raw data. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.  2562–2571, 2019. URL https://api.semanticscholar.org/CorpusID:208267630.
  3. Tensorf: Tensorial radiance fields. ArXiv, abs/2203.09517, 2022. URL https://api.semanticscholar.org/CorpusID:247519170.
  4. Dictionary fields: Learning a neural basis decomposition. ACM Transactions on Graphics (TOG), 42:1 – 12, 2023a. URL https://api.semanticscholar.org/CorpusID:260167858.
  5. Factor fields: A unified framework for neural fields and beyond. ArXiv, abs/2302.01226, 2023b. URL https://api.semanticscholar.org/CorpusID:256503583.
  6. Nerv: Neural representations for videos. In Neural Information Processing Systems, 2021. URL https://api.semanticscholar.org/CorpusID:239885704.
  7. Learning continuous image representation with local implicit image function. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.  8624–8634, 2020. URL https://api.semanticscholar.org/CorpusID:229221619.
  8. Learning implicit fields for generative shape modeling. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.  5932–5941, 2018. URL https://api.semanticscholar.org/CorpusID:54457478.
  9. Activation functions in deep learning: A comprehensive survey and benchmark. Neurocomputing, 503:92–108, 2021. URL https://api.semanticscholar.org/CorpusID:250089226.
  10. Implicit geometric regularization for learning shapes. ArXiv, abs/2002.10099, 2020. URL https://api.semanticscholar.org/CorpusID:211259068.
  11. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. 2015 IEEE International Conference on Computer Vision (ICCV), pp.  1026–1034, 2015. URL https://api.semanticscholar.org/CorpusID:13740328.
  12. Deep learning with s-shaped rectified linear activation units. In AAAI Conference on Artificial Intelligence, 2015. URL https://api.semanticscholar.org/CorpusID:10520992.
  13. Performance analysis of various activation functions in generalized mlp architectures of neural networks. 2011. URL https://api.semanticscholar.org/CorpusID:174791561.
  14. Neural 3d mesh renderer. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.  3907–3916, 2017. URL https://api.semanticscholar.org/CorpusID:32389979.
  15. Siamese siren: Audio compression with implicit neural representations. ArXiv, abs/2306.12957, 2023. URL https://api.semanticscholar.org/CorpusID:259224407.
  16. Nonlinear signal processing using neural networks: Prediction and system modelling. 1987. URL https://api.semanticscholar.org/CorpusID:60720876.
  17. Improving deep neural network with multiple parametric exponential linear units. ArXiv, abs/1606.00305, 2016. URL https://api.semanticscholar.org/CorpusID:9248703.
  18. Fourier neural operator for parametric partial differential equations. ArXiv, abs/2010.08895, 2020. URL https://api.semanticscholar.org/CorpusID:224705257.
  19. Acorn. ACM Transactions on Graphics (TOG), 40:1 – 13, 2021. URL https://api.semanticscholar.org/CorpusID:233864500.
  20. Occupancy networks: Learning 3d reconstruction in function space. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.  4455–4465, 2018. URL https://api.semanticscholar.org/CorpusID:54465161.
  21. Implicit surface representations as layers in neural networks. 2019 IEEE/CVF International Conference on Computer Vision (ICCV), pp.  4742–4751, 2019. URL https://api.semanticscholar.org/CorpusID:207985696.
  22. Nerf. Communications of the ACM, 65:99 – 106, 2020. URL https://api.semanticscholar.org/CorpusID:213175590.
  23. Instant neural graphics primitives with a multiresolution hash encoding. ACM Transactions on Graphics (TOG), 41:1 – 15, 2022. URL https://api.semanticscholar.org/CorpusID:246016186.
  24. Texture fields: Learning texture representations in function space. 2019 IEEE/CVF International Conference on Computer Vision (ICCV), pp.  4530–4539, 2019. URL https://api.semanticscholar.org/CorpusID:158046789.
  25. Taming the waves: sine as activation function in deep neural networks. 2017. URL https://api.semanticscholar.org/CorpusID:126004626.
  26. Deepsdf: Learning continuous signed distance functions for shape representation. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.  165–174, 2019. URL https://api.semanticscholar.org/CorpusID:58007025.
  27. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys., 378:686–707, 2019. URL https://api.semanticscholar.org/CorpusID:57379996.
  28. Searching for activation functions. ArXiv, abs/1710.05941, 2018. URL https://api.semanticscholar.org/CorpusID:10919244.
  29. Beyond periodicity: Towards a unifying framework for activations in coordinate-mlps. ArXiv, abs/2111.15135, 2021. URL https://api.semanticscholar.org/CorpusID:244729036.
  30. Implicit neural representations with periodic activation functions. ArXiv, abs/2006.09661, 2020. URL https://api.semanticscholar.org/CorpusID:219720931.
  31. Neural networks with periodic and monotonic activation functions: a comparative study in classification problems. 1999. URL https://api.semanticscholar.org/CorpusID:122973551.
  32. Fourier features let networks learn high frequency functions in low dimensional domains. ArXiv, abs/2006.10739, 2020. URL https://api.semanticscholar.org/CorpusID:219791950.
  33. Advances in neural rendering. Computer Graphics Forum, 41, 2021. URL https://api.semanticscholar.org/CorpusID:236162433.
  34. State of the art on neural rendering. Computer Graphics Forum, 39, 2020. URL https://api.semanticscholar.org/CorpusID:215416317.
  35. Handwritten digit recognition using multilayer feedforward neural networks with periodic and monotonic activation functions. Object recognition supported by user interaction for service robots, 3:106–109 vol.3, 2002. URL https://api.semanticscholar.org/CorpusID:16161892.
  36. A survey on deep geometry learning: From a representation perspective. Computational Visual Media, 6:113 – 133, 2020. URL https://api.semanticscholar.org/CorpusID:211171854.
  37. Neural fields in visual computing and beyond. Computer Graphics Forum, 41, 2021. URL https://api.semanticscholar.org/CorpusID:244478496.
Citations (4)
View on Semantic Scholar

Summary

No one has generated a summary of this paper yet.

Sign Up to Summarize

Paper to Video (Beta)

No one has generated a video about this paper yet.

Sign Up to Generate All Videos Subscribe on YouTube

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Sign Up to Generate

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Generate Now

Continue Learning

We haven't generated follow-up questions for this paper yet.

Generate Now

Collections

Sign up for free to add this paper to one or more collections.

Sign Up

Tweets

Sign up for free to view the 1 tweet with 0 likes about this paper.

Sign Up for Free