DynaDrag: Dynamic Drag-Style Image Editing by Motion Prediction

TL;DR

DynaDrag introduces a novel predict-and-move framework for pixel-level image editing that improves accuracy and editability by iteratively predicting and adjusting handle points, outperforming previous methods.

Contribution

It is the first to implement a predict-and-move framework with dynamic handle point adjustment for drag-style image editing.

Findings

Outperforms previous methods on face and human datasets.

Reduces issues of tracking failure and ambiguous tracking.

Enhances editability through dynamic handle point adjustment.

Abstract

To achieve pixel-level image manipulation, drag-style image editing which edits images using points or trajectories as conditions is attracting widespread attention. Most previous methods follow move-and-track framework, in which miss tracking and ambiguous tracking are unavoidable challenging issues. Other methods under different frameworks suffer from various problems like the huge gap between source image and target edited image as well as unreasonable intermediate point which can lead to low editability. To avoid these problems, we propose DynaDrag, the first dragging method under predict-and-move framework. In DynaDrag, Motion Prediction and Motion Supervision are performed iteratively. In each iteration, Motion Prediction first predicts where the handle points should move, and then Motion Supervision drags them accordingly. We also propose to dynamically adjust the valid handle points to further improve the performance. Experiments on face and human datasets showcase the superiority over previous works.