OPAL-Net: A Generative Model for Part-based Object Layout Generation

TL;DR

OPAL-Net is a hierarchical generative model that creates detailed, category-aware object layouts using a unified approach with graph and recurrent neural networks, trained on PASCAL-Parts.

Contribution

It introduces a novel hierarchical architecture combining GCNs, RNNs, and VAEs for flexible, diverse part-based object layout generation across multiple categories.

Findings

Generated layouts are diverse and category-aware.

OPAL-Net outperforms baseline models in evaluation scores.

The model demonstrates versatility across object categories.

Abstract

We propose OPAL-Net, a novel hierarchical architecture for part-based layout generation of objects from multiple categories using a single unified model. We adopt a coarse-to-fine strategy involving semantically conditioned autoregressive generation of bounding box layouts and pixel-level part layouts for objects. We use Graph Convolutional Networks, Deep Recurrent Networks along with custom-designed Conditional Variational Autoencoders to enable flexible, diverse and category-aware generation of object layouts. We train OPAL-Net on PASCAL-Parts dataset. The generated samples and corresponding evaluation scores demonstrate the versatility of OPAL-Net compared to ablative variants and baselines.