Search results
Jul 20, 2022 · The official repo for CM-GAN (Cascaded Modulation GAN) for Image Inpainting. We introduce a new cascaded modulation design that cascades global modulation with spatial adaptive modulation for better hole filling. We also introduce an object-aware training scheme to facilitate better object removal.
Mar 22, 2022 · We propose cascaded modulation GAN (CM-GAN), a new network design consisting of an encoder with Fourier convolution blocks that extract multi-scale feature representations from the input image with holes and a dual-stream decoder with a novel cascaded global-spatial modulation block at each scale level. In each decoder block, global modulation ...
- arXiv:2203.11947 [cs.CV]
- 32 pages, 19 figures
We demonstrate the setup by combing a full body GAN with a dedicated high-quality face GAN to produce plausible-looking humans. We evaluate our results with quantitative metrics and user studies. Learn More
Mar 14, 2022 · We demonstrate the setup by combining a full body GAN with a dedicated high-quality face GAN to produce plausible-looking humans. We evaluate our results with quantitative metrics and user studies.
– Cascaded Modulation GAN, a new inpainting network architecture formed by a masked image encoder with Fourier convolution blocks and a cascaded global-spatial modulation-based decoder. – An object-aware mask generation scheme preventing the model from gener-ating new objects inside holes and mimicking realistic inpainting use cases. – A ...
Instead of modeling this complex domain with a single GAN, we propose a novel method to combine multiple pretrained GANs, where one GAN generates a global canvas (e.g., human body) and a set of specialized GANs, or insets, focus on different parts (e.g., faces, shoes) that can be seamlessly inserted onto the global canvas.
Oct 21, 2022 · To seek a better way to inject global context into the missing region in inpainting, we investigate a new modulation scheme by cascading global and spatial modulations. We propose Cascaded Modulation GAN (CM-GAN), a new generative network that can synthesize better holistic structure and local details, cf. Figs 1 and 5.
