KwaiMind improves commercial image editing accuracy and appeal
KwaiMind Technical Report
Computer Vision and Pattern Recognition
Summary
Editing product images for online stores is hard because the product needs to look right, text must be clear, and customers want to be attracted. The authors designed KwaiMind, a smart system trained on millions of editing examples and guided by how likely customers are to click the edited images. It works better than other open-source tools on many tests and even leads to more clicks in real online experiments. This shows that using special data and customer feedback helps make better product images.
What this means in practice
- •For e-commerce platform teams: Improve product main images by selecting edits that increase customer click-through rates during online shopping.$Commercial implications: Enables a product image optimization service that boosts clicks on e-commerce listings through AI-driven editing and selection.
- •For digital marketing teams: Create visually appealing product images that keep brand identity and render text clearly for advertising campaigns.
Authors
Junlong Wu, Zijun Li, Yuting Hu, Jia Sun, Pengcheng Wei, Yimin Zhou, Honglie Wang, Huaiqing Wang, Dewen Fan, Fei Zuo, Haixuan Gao, Lihui Peng, Tingxuan She, Yuqing Li, Boheng Zhang, Fan Yang, Wenwu Ou
Abstract
Commercial image editing requires product identity preservation, accurate text rendering, and user appeal alongside general editing quality. We present KwaiMind, an image editing system combining general capabilities with e-commerce specialization. An agent-based data engine maintains approximately 1.8 million high-quality editing pairs. Built on a multimodal diffusion transformer, KwaiMind undergoes continued pre-training and supervised fine-tuning, followed by preference optimization and online reinforcement learning. A general-purpose vision-language judge and specialized rewards for click-through rate (CTR), text rendering, and product consistency guide specialized policies, which are consolidated through on-policy distillation. We introduce Ecom-Bench, covering 11 commercial editing tasks with task-specific visual evaluation and CTR-based ranking. KwaiMind achieves the strongest overall scores among evaluated open-source editors on ImgEdit, GEdit, both language splits of REDEdit, and Ecom-Bench visual quality, and the highest aggregate CTR ranking score among compared systems. Offline, CTR-guided optimization increases the proportion of generated images whose predicted CTR exceeds that of the original product image from 12.16% to 37.41%. In an online A/B experiment, CTR-based selection of product main images yields an approximately 2.44% relative increase in actual CTR. These results demonstrate the value of domain-specific data and reward-driven alignment for commercial image editing.