Proximal Deep Structured Models

Authors: Shenlong Wang, Sanja Fidler, Raquel Urtasun

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the effectiveness of our approach in the tasks of image denoising, depth refinement and optical flow estimation.
Researcher Affiliation Academia Shenlong Wang University of Toronto slwang@cs.toronto.edu Sanja Fidler University of Toronto fidler@cs.toronto.edu Raquel Urtasun University of Toronto urtasun@cs.toronto.edu
Pseudocode Yes Figure 2: Algorithm for learning proximal deep structured models. (Contains a numbered list of steps)
Open Source Code No The paper does not provide an explicit statement or link for the open-source code of the described methodology.
Open Datasets Yes We use the BSDS image dataset [23]. (for image denoising) We conduct the depth refinement experiment on the 7 Scenes dataset [25]. (for depth refinement) We evaluate the task of optical flow estimation on the Flying Chairs dataset [11]. (for optical flow)
Dataset Splits No The paper mentions using training and testing subsets but does not explicitly provide specific details for a validation dataset split.
Hardware Specification Yes Our experiments are conducted on a Xeon 3.2 Ghz machine with a Titan X GPU.
Software Dependencies Yes We employ mxnet [4] with CUDNNv4 acceleration to implement the networks
Experiment Setup Yes We use mean square error as the loss function and set a weight decay strength of 0.0004 for all settings. MSRA initialization [16] is used for the convolution parameters and the initial gradient step for each iteration is set to be 0.02. We use adam [19] with a learning rate of t = 0.02 and hyper-parameters β1 = 0.9 and β2 = 0.999 as in Kingma et al. [19]. The learning rate is divided by 2 every 50 epoch, and we use a mini-batch size of 32.