SKU-DiT
An image generation model based on the Diffusion Transformer architecture. It supports joint conditioning on three types of input: white-background product images, logo-cropped images and structured layouts, and is optimized specifically for e-commerce main-image generation.
Timeline (last 90 days)
SKU-DiT accepts three types of conditional input simultaneously: white-background product images, logo crop images, and structured layout information, corresponding to the three constraints of subject fidelity, logo integrity, and character-level text accuracy.
SKU-DiT adopts the Reinforce-Distill-Reinforce (RDR) progressive training strategy: first exploring the generation space through reinforcement learning, then compressing the model's capabilities through distillation, and finally reinforcing again to restore and improve performance.
All Facts (2)
SKU-DiT accepts three types of conditional input simultaneously: white-background product images, logo crop images, and structured layout information, corresponding to the three constraints of subject fidelity, logo integrity, and character-level text accuracy.
50%UnverifiedSKU-DiT adopts the Reinforce-Distill-Reinforce (RDR) progressive training strategy: first exploring the generation space through reinforcement learning, then compressing the model's capabilities through distillation, and finally reinforcing again to restore and improve performance.
50%