All Articles

Hot Papers 2021-01-27

1. Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes

Towaki Takikawa, Joey Litalien, Kangxue Yin, Karsten Kreis, Charles Loop, Derek Nowrouzezahrai, Alec Jacobson, Morgan McGuire, Sanja Fidler

  • retweets: 4371, favorites: 150 (01/28/2021 09:25:41)
  • links: abs | pdf
  • cs.CV | cs.GR

Neural signed distance functions (SDFs) are emerging as an effective representation for 3D shapes. State-of-the-art methods typically encode the SDF with a large, fixed-size neural network to approximate complex shapes with implicit surfaces. Rendering with these large networks is, however, computationally expensive since it requires many forward passes through the network for every pixel, making these representations impractical for real-time graphics. We introduce an efficient neural representation that, for the first time, enables real-time rendering of high-fidelity neural SDFs, while achieving state-of-the-art geometry reconstruction quality. We represent implicit surfaces using an octree-based feature volume which adaptively fits shapes with multiple discrete levels of detail (LODs), and enables continuous LOD with SDF interpolation. We further develop an efficient algorithm to directly render our novel neural SDF representation in real-time by querying only the necessary LODs with sparse octree traversal. We show that our representation is 2-3 orders of magnitude more efficient in terms of rendering speed compared to previous works. Furthermore, it produces state-of-the-art reconstruction quality for complex shapes under both 3D geometric and 2D image-space metrics.

2. English Machine Reading Comprehension Datasets: A Survey

Daria Dzendzik, Carl Vogel, Jennifer Foster

  • retweets: 1562, favorites: 170 (01/28/2021 09:25:41)
  • links: abs | pdf
  • cs.CL

This paper surveys 54 English Machine Reading Comprehension datasets, with a view to providing a convenient resource for other researchers interested in this problem. We categorize the datasets according to their question and answer form and compare them across various dimensions including size, vocabulary, data source, method of creation, human performance level, and first question word. Our analysis reveals that Wikipedia is by far the most common data source and that there is a relative lack of why, when, and where questions across datasets.

3. Modeling Assumptions Clash with the Real World: Transparency, Equity, and Community Challenges for Student Assignment Algorithms

Samantha Robertson, Tonya Nguyen, Niloufar Salehi

Across the United States, a growing number of school districts are turning to matching algorithms to assign students to public schools. The designers of these algorithms aimed to promote values such as transparency, equity, and community in the process. However, school districts have encountered practical challenges in their deployment. In fact, San Francisco Unified School District voted to stop using and completely redesign their student assignment algorithm because it was not promoting educational equity in practice. We analyze this system using a Value Sensitive Design approach and find that one reason values are not met in practice is that the system relies on modeling assumptions about families’ priorities, constraints, and goals that clash with the real world. These assumptions overlook the complex barriers to ideal participation that many families face, particularly because of socioeconomic inequalities. We argue that direct, ongoing engagement with stakeholders is central to aligning algorithmic values with real world conditions. In doing so we must broaden how we evaluate algorithms while recognizing the limitations of purely algorithmic solutions in addressing complex socio-political problems.

4. Bibliometric assessment of national scientific journals

Henk F. Moed, Felix de Moya-Anegon, Vicente Guerrero-Bote, Carmen Lopez-Illescas, Myroslava Hladchenko

  • retweets: 196, favorites: 18 (01/28/2021 09:25:42)
  • links: abs | pdf
  • cs.DL

Nationally oriented scientific-scholarly journals are considered from a methodological-informetric viewpoint, analysing data extracted from Scimago Journal Rank based on Scopus. An operational definition is proposed of a journal’s degree of national orientation based on the geographical distribution of its publishing or citing authors, and the role of international collaboration and a country’s total publication output. A comprehensive analysis is presented of trends up until 2019 in national orientation and citation impact of national journals entering Scopus, extending outcomes in earlier studies. A method to analyse national journals of given countries is applied to the set of former USSR republics and Eastern and Central European states which were under socialism, distinguishing between domestic and foreign national journals. The possible influence is highlighted of factors related to a journal’s access status, publication language and subject field, international scientific migration and collaboration, database coverage policies, the size of a national research community, historical-political factors and national research assessment and funding policies.

5. Deep Burst Super-Resolution

Goutam Bhat, Martin Danelljan, Luc Van Gool, Radu Timofte

  • retweets: 90, favorites: 57 (01/28/2021 09:25:42)
  • links: abs | pdf
  • cs.CV

While single-image super-resolution (SISR) has attracted substantial interest in recent years, the proposed approaches are limited to learning image priors in order to add high frequency details. In contrast, multi-frame super-resolution (MFSR) offers the possibility of reconstructing rich details by combining signal information from multiple shifted images. This key advantage, along with the increasing popularity of burst photography, have made MFSR an important problem for real-world applications. We propose a novel architecture for the burst super-resolution task. Our network takes multiple noisy RAW images as input, and generates a denoised, super-resolved RGB image as output. This is achieved by explicitly aligning deep embeddings of the input frames using pixel-wise optical flow. The information from all frames are then adaptively merged using an attention-based fusion module. In order to enable training and evaluation on real-world data, we additionally introduce the BurstSR dataset, consisting of smartphone bursts and high-resolution DSLR ground-truth. We perform comprehensive experimental analysis, demonstrating the effectiveness of the proposed architecture.

6. CPTR: Full Transformer Network for Image Captioning

Wei Liu, Sihan Chen, Longteng Guo, Xinxin Zhu, Jing Liu

  • retweets: 56, favorites: 64 (01/28/2021 09:25:42)
  • links: abs | pdf
  • cs.CV

In this paper, we consider the image captioning task from a new sequence-to-sequence prediction perspective and propose Caption TransformeR (CPTR) which takes the sequentialized raw images as the input to Transformer. Compared to the “CNN+Transformer” design paradigm, our model can model global context at every encoder layer from the beginning and is totally convolution-free. Extensive experiments demonstrate the effectiveness of the proposed model and we surpass the conventional “CNN+Transformer” methods on the MSCOCO dataset. Besides, we provide detailed visualizations of the self-attention between patches in the encoder and the “words-to-patches” attention in the decoder thanks to the full Transformer architecture.

7. Exploring Transitivity in Neural NLI Models through Veridicality

Hitomi Yanaka, Koji Mineshima, Kentaro Inui

  • retweets: 56, favorites: 22 (01/28/2021 09:25:42)
  • links: abs | pdf
  • cs.CL

Despite the recent success of deep neural networks in natural language processing, the extent to which they can demonstrate human-like generalization capacities for natural language understanding remains unclear. We explore this issue in the domain of natural language inference (NLI), focusing on the transitivity of inference relations, a fundamental property for systematically drawing inferences. A model capturing transitivity can compose basic inference patterns and draw new inferences. We introduce an analysis method using synthetic and naturalistic NLI datasets involving clause-embedding verbs to evaluate whether models can perform transitivity inferences composed of veridical inferences and arbitrary inference types. We find that current NLI models do not perform consistently well on transitivity inference tasks, suggesting that they lack the generalization capacity for drawing composite inferences from provided training examples. The data and code for our analysis are publicly available at https://github.com/verypluming/transitivity.

Published 28 Jan 2021

ML Lead at Beatrust. (https://beatrust.com)Tatsuya Shirakawa on Twitter