INNER CODE UNIT · Python

LocalBlend

google/break-a-scene · ptp_utils.py:173

class LocalBlend:
    def __call__(self, x_t, attention_store):
        k = 1
        maps = attention_store["down_cross"][2:4] + attention_store["up_cross"][:3]
        maps = [
            item.reshape(self.alpha_layers.shape[0], -1, 1, 16, 16, self.max_num_words)
            for item in maps
        ]
        maps = torch.cat(maps, dim=1)
        maps = (maps * self.alpha_layers).sum(-1).mean(1)
        mask = nnf.max_pool2d(maps, (k * 2 + 1, k * 2 + 1), (1, 1), padding=(k, k))
        mask = nnf.interpolate(mask, size=(x_t.shape[2:]))
        mask = mask / mask.max(2, keepdims=True)[0].max(3, keepdims=True)[0]
        mask = mask.gt(self.threshold)
        mask = (mask[:1] + mask[1:]).float()
        x_t = x_t[:1] + mask * (x_t - x_t[:1])
        return x_t

View source record →

📰 Research Paper
Loading…
⏳ Fetching content…