Probabilistic Neural Programmed Networks for Scene Generation

Authors: Zhiwei Deng, Jiacheng Chen, YIFANG FU, Greg Mori

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

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate PNP-Net on the task of text to scene generation across a series of datasets and experimental settings with various complexity.The results are summarized in Tab. 5b.
Researcher Affiliation Academia Zhiwei Deng, Jiacheng Chen, Yifang Fu, Greg Mori Simon Fraser University {zhiweid, jca348, yifangf}@sfu.ca, mori@cs.sfu.ca
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes More details can be found in our project repository2 where we release the code for model training/evaluating and dataset generation. Footnote 2: https://github.com/Lucas2012/Probabilistic Neural Programmed Network
Open Datasets Yes We create a Color MNIST dataset which contains images with up to two digits in an image. We create a CLEVR-G dataset which contains 10000 64x64 training images and 10000 testing images. More details can be found in our project repository2 where we release the code for model training/evaluating and dataset generation.
Dataset Splits No The paper specifies '8000 training and 8000 test images' for Color MNIST and '10000 training images and 10000 testing images' for CLEVR-G, but does not explicitly mention a validation split or its size.
Hardware Specification No The paper does not provide specific details on hardware used, such as GPU/CPU models or memory.
Software Dependencies No The paper mentions 'Adamax [36]' for optimization but does not provide specific version numbers for any software dependencies or libraries.
Experiment Setup Yes The hidden dimension size is 160 and batchnorm layer is added to stabilize training. For latent size for appearance distribution, we set the latent size to be 64 dimensions with height and width as 16. For location/scale latent distribution, we use 8 dimensions. The learning rate is set as 0.0001 and the model is optimized by Adamax [36].