PartStickers: Generating Parts of Objects for Rapid Prototyping
University of Colorado Boulder
University of Colorado Boulder
Design prototyping involves creating mockups of products or concepts to gather feedback and iterate on ideas. While prototyping often requires specific parts of objects, such as when constructing a novel creature for a video game, existing text-to-image methods tend to only generate entire objects. To address this, we propose a novel task and method of ``part sticker generation", which entails generating an isolated part of an object on a neutral background. Experiments demonstrate our method outperforms state-of-the-art baselines with respect to realism and text alignment, while preserving object-level generation capabilities.
We design a pipeline to support our task of part sticker generation, by generating a single part specified in a text prompt that is `pasted' on a neutral, gray background. The pipeline consists of two key steps. First, given an image with associated object and part segmentation masks, we localize each desired part and place it in the center of a gray background. Second, for each localized image region, we pair it with a prompt describing the region. All resulting image-prompt pairs are then used to train a text-to-image diffusion model. We hypothesize that such a pipeline would enable a trained model to generate diverse renderings for each part type because existing datasets offer many diverse part segmentations, including of the same types observed across different objects (e.g., an eye can be found in a frog and dog).
Qualitative results showing examples of generated images given text prompts (left) and the average image of 100 generated samples from a given method (left). Overall, we observe that PartStickers is the only method capable of consistently generating only the requested part on a neutral background with a high degree of realism. The bottom three rows represent out-of-distribution scenarios for PartStickers: generation of an object and two out-of-distribution parts. (SD stands for Stable Diffusion).
@inproceedings{zhou2025partstickers, author = {Mo, Zhou and Josh, Myers-Dean and Danna, Gurari}, title = {PartStickers: Generating Parts of Objects for Rapid Prototyping}, booktitle = {Proceedings of the CVEU Workshop at CVPR 2025}, year = {2025}, publisher = {IEEE}, address = {Nashville, TN, USA}, }