[2006.09662] MetaSDF: Meta-learning Signed Distance Functionsopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

Neural implicit shape representations are an emerging paradigm that offers many potential benefits over conventional discrete representations, including memory efficiency at a high spatial resolution. Generalizing across shapes with such neural implicit representations amounts to learning priors over the respective function space and enables geometry reconstruction from partial or noisy observations. Existing generalization methods rely on conditioning a neural network on a low-dimensional latent code that is either regressed by an encoder or jointly optimized in the auto-decoder framework. Here, we formalize learning of a shape space as a meta-learning problem and leverage gradient-based meta-learning algorithms to solve this task. We demonstrate that this approach performs on par with auto-decoder based approaches while being an order of magnitude faster at test-time inference. We further demonstrate that the proposed gradient-based method outperforms encoder-decoder based methods th

2 mentions: @hillbig@hillbig
Date: 2020/11/18 23:22

Referring Tweets

@hillbig MetaSDFは(部分)観測からのSDFによる形状推定をメタ学習(MAML)で実現。事前情報の利用として画像を符号化して得た潜在ベクトルで条件付けしたりハイパーネットワークを使う場合に比べ精度を大きく改善、収束(SDFによる形状獲得)速度が8→0.4秒と20倍近く速くなった t.co/d5z78MQKVo
@hillbig MetaSDF learns priors over shapes by meta-learning (MAML). Unlike the approaches using latent code or hyper-network, meta-learning can reconstruct a shape (estimate SDF for given shape) significantly faster, requires only 5 update steps in 0.4 seconds. t.co/d5z78MQKVo

Related Entries

Read more Caveats for information bottleneck in deterministic scenarios | OpenReview
0 users, 2 mentions 2019/02/12 02:17
Read more Gradient Descent Maximizes the Margin of Homogeneous Neural Networks | OpenReview
0 users, 2 mentions 2019/11/09 02:20
Read more [1905.05894] Online Normalization for Training Neural Networks
0 users, 2 mentions 2019/12/17 00:51
Read more [2002.07942] Source Separation with Deep Generative Priorscontact arXivarXiv Twitter
0 users, 2 mentions 2020/02/25 18:51
Read more [2001.06499] Temporal Interlacing Networkcontact arXivarXiv Twitter
0 users, 2 mentions 2020/04/13 00:52
Read more DIST - Differentiable Sphere Tracing
0 users, 2 mentions 2020/05/07 00:52