Context-aware Synthesis and Placement of Object Instances

Authors: Donghoon Lee, Sifei Liu, Jinwei Gu, Ming-Yu Liu, Ming-Hsuan Yang, Jan Kautz

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experimental validations with comparisons to strong baselines to verify the effectiveness of the proposed network. Experimental results on benchmark datasets show that the proposed algorithm can synthesize plausible and diverse pairs of location and appearance for new objects.
Researcher Affiliation Collaboration Donghoon Lee1,2 , Sifei Liu3, Jinwei Gu3, Ming-Yu Liu3, Ming-Hsuan Yang2,4, Jan Kautz3 donghoon.lee@rllab.snu.ac.kr {sifeil, jinweig, mingyul}@nvidia.com mhyang@ucmerced.edu jkautz@nvidia.com 1Seoul National University, 2Google Cloud AI, 3NVIDIA, 4University of California at Merced
Pseudocode No The paper does not contain any pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes Code is available at https: //github.com/NVlabs/Instance_Insertion.
Open Datasets Yes we use datasets that provide both semantic and instance-level annotations, e.g., Cityscapes [3].
Dataset Splits No The paper mentions using the Cityscapes dataset but does not provide specific details on training, validation, or test splits (e.g., percentages or sample counts).
Hardware Specification No The paper does not explicitly describe the hardware used to run its experiments. It does not mention specific GPU models, CPU models, or cloud computing instance types.
Software Dependencies No The paper states that
Experiment Setup Yes The batch size is set to 1 and instance normalization is used instead of batch normalization. We use transposed convolutional layers with 32 as a base number of filters to generate the shape, while we use 16 for convolutional layers in the discriminator.