INNER CODE UNIT · Python

mapped_img

wenhaochai/StableVideo · app.py:213

                mapped_img = 2. * mapped_img - 1.
                latent = self.canny_model.get_first_stage_encoding(self.canny_model.encode_first_stage(mapped_img))
                
                t_enc = int(ddim_steps * s)
                latent = self.canny_ddim_sampler.stochastic_encode(latent, torch.tensor([t_enc]).to("cuda"))
                samples = self.canny_ddim_sampler.decode(x_latent=latent, 
                                                         cond=cond, 
                                                         t_start=t_enc,
                                                         unconditional_guidance_scale=scale,
                                                         unconditional_conditioning=un_cond)

            x_samples = self.canny_model.decode_first_stage(samples)
            result = (einops.rearrange(x_samples, 'b c h w -> b h w c') * 127.5 + 127.5).cpu().numpy().clip(0, 255).astype(np.uint8)
            result = Image.fromarray(result[0])
            
            result = result.resize(current_img.size)
            result = transforms.ToTensor()(result)
            # times alpha

View source record →

📰 Research Paper
Loading…
⏳ Fetching content…