PartCraft: Crafting Creative Objects by Parts

TL;DR

PartCraft introduces a novel method for fine-grained, part-based visual generation in AI, enabling users to craft highly customized objects by selecting and manipulating object parts, with improved fidelity and generalization.

Contribution

This work pioneers part-based control in generative visual AI, combining unsupervised part parsing, entropy-based attention loss, and a bottleneck encoder for enhanced fidelity and generalization.

Findings

Demonstrates high-quality, customized object generation with part control.

Achieves better generalization to novel part compositions.

Code released for reproducibility and further research.

Abstract

This paper propels creative control in generative visual AI by allowing users to "select". Departing from traditional text or sketch-based methods, we for the first time allow users to choose visual concepts by parts for their creative endeavors. The outcome is fine-grained generation that precisely captures selected visual concepts, ensuring a holistically faithful and plausible result. To achieve this, we first parse objects into parts through unsupervised feature clustering. Then, we encode parts into text tokens and introduce an entropy-based normalized attention loss that operates on them. This loss design enables our model to learn generic prior topology knowledge about object's part composition, and further generalize to novel part compositions to ensure the generation looks holistically faithful. Lastly, we employ a bottleneck encoder to project the part tokens. This not only enhances fidelity but also accelerates learning, by leveraging shared knowledge and facilitating information exchange among instances. Visual results in the paper and supplementary material showcase the compelling power of PartCraft in crafting highly customized, innovative creations, exemplified by the "charming" and creative birds. Code is released at https://github.com/kamwoh/partcraft.