Learning a Multi-View Stereo Machine

Authors: Abhishek Kar, Christian Häne, Jitendra Malik

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We thoroughly evaluate our approach on the Shape Net dataset and demonstrate the benefits over classical approaches and recent learning based methods. In this section, we demonstrate the ability of LSMs to learn 3D shape reconstruction in a geometrically accurate manner. First, we present quantitative results for V-LSMs on the Shape Net dataset [3] and compare it to various baselines, both classical and learning based.
Researcher Affiliation Academia Abhishek Kar UC Berkeley akar@berkeley.edu Christian Häne UC Berkeley chaene@berkeley.edu Jitendra Malik UC Berkeley malik@berkeley.edu
Pseudocode No The paper describes the system architecture and its components in narrative text, but it does not include any formal pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any statement about making the source code available or provide a link to a code repository.
Open Datasets Yes We use the synthetic Shape Net dataset [3] to generate posed image-sets, ground truth 3D occupancy maps and depth maps for all our experiments. A. X. Chang, T. Funkhouser, L. Guibas, P. Hanrahan, Q. Huang, Z. Li, S. Savarese, M. Savva, S. Song, H. Su, J. Xiao, L. Yi, and F. Yu. Shapenet: An information-rich 3d model repository. ar Xiv preprint ar Xiv:1512.03012, 2015.
Dataset Splits Yes More specifically, we use a subset of 13 major categories (same as [5]) containing around 44k 3D models resized to lie within the unit cube centered at the origin with a train/val/test split of [0.7, 0.1, 0.2].
Hardware Specification Yes The projection and unprojection operations are trivially implemented on the GPU with batched matrix multiplications and bilinear/nearest sampling enabling inference at around 30 models/sec on a GTX 1080Ti.
Software Dependencies No We implemented our networks in Tensorflow and trained both the variants of LSMs for 100k iterations using Adam. The paper mentions TensorFlow and Adam, but does not provide specific version numbers for these software components.
Experiment Setup Yes We use 224 224 images to train LSMs with a shape batch size of 4 and 4 views per shape. Our world grid is at a resolution of 323. We implemented our networks in Tensorflow and trained both the variants of LSMs for 100k iterations using Adam.