IR-GAN: Image Manipulation with Linguistic Instruction by Increment Reasoning
This addresses a challenge in multimodal conditional generation for image manipulation, but it is incremental as it builds on existing GAN-based methods.
The paper tackles the problem of partial inconsistency between images and linguistic instructions in conditional image generation by proposing IR-GAN, which reasons about visual and semantic increments to improve consistency, achieving effective results as shown in experiments on two datasets.
Conditional image generation is an active research topic including text2image and image translation. Recently image manipulation with linguistic instruction brings new challenges of multimodal conditional generation. However, traditional conditional image generation models mainly focus on generating high-quality and visually realistic images, and lack resolving the partial consistency between image and instruction. To address this issue, we propose an Increment Reasoning Generative Adversarial Network (IR-GAN), which aims to reason the consistency between visual increment in images and semantic increment in instructions. First, we introduce the word-level and instruction-level instruction encoders to learn user's intention from history-correlated instructions as semantic increment. Second, we embed the representation of semantic increment into that of source image for generating target image, where source image plays the role of referring auxiliary. Finally, we propose a reasoning discriminator to measure the consistency between visual increment and semantic increment, which purifies user's intention and guarantees the good logic of generated target image. Extensive experiments and visualization conducted on two datasets show the effectiveness of IR-GAN.