NeurVPS: Neural Vanishing Point Scanning via Conic Convolution

Authors: Yichao Zhou, Haozhi Qi, Jingwei Huang, Yi Ma

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our extensive experiments on both synthetic and real-world datasets show that the proposed operator significantly improves the performance of vanishing point detection over traditional methods.
Researcher Affiliation Academia Yichao Zhou UC Berkeley zyc@berkeley.edu Haozhi Qi UC Berkeley hqi@berkeley.edu Jingwei Huang Standford University jingweih@stanford.edu Yi Ma UC Berkeley yima@eecs.berkeley.edu
Pseudocode No The paper describes the algorithm and its components in text and diagrams but does not include formal pseudocode blocks or algorithm listings.
Open Source Code Yes The code and dataset have been made publicly available at https://github.com/zhou13/neurvps.
Open Datasets Yes We conduct experiments on both synthetic [49] and real-world [50, 13] datasets. ... Natural Scene [50]. ... Scan Net [13]. ... SU3 Wireframe [49].
Dataset Splits Yes Natural Scene [50]: We divide them into 2,000 training images and 275 test images randomly. ... Scan Net [13]: There are 266,844 training images. We randomly sample 500 images from the validation set as our test set. ... SU3 Wireframe [49]: It contains 22,500 training images and 500 validation images.
Hardware Specification Yes All experiments are conducted on two NVIDIA RTX 2080Ti GPUs
Software Dependencies No We implement the conic convolution operator in Py Torch by modifying the im2col + GEMM function according to Equation (3)... The paper mentions PyTorch but does not specify a version number or other software dependencies with their versions.
Experiment Setup Yes Input images are resized to 512 512. During training, the Adam optimizer [25] is used. Learning rate and weight decay are set to be 4 10 4 and 1 10 5, respectively. ... For synthetic data [49], we train 30 epochs and reduce the learning rate by 10 at the 24-th epoch. ... For the Natural Scene dataset, we train the model for 100 epochs and decay the learning rate at 60-th epoch. For Scan Net [13], we train the model for 3 epochs. ... We set Nd = 64 and use RSU3 = 5, RNS = 4, and RSN = 3 in the coarse-to-fine inference for the SU3 dataset, the Natural Scene dataset, and the Scan Net dataset, respectively.