import warnings from pathlib import Path from typing import Callable, List, Optional import numpy as np import torch import torch.nn.functional as F from omegaconf import OmegaConf from torch import nn from torch.utils.checkpoint import checkpoint from ...settings import DATA_PATH from ..utils.losses import NLLLoss from ..utils.metrics import matcher_metrics FLASH_AVAILABLE = hasattr(F, "scaled_dot_product_attention") torch.backends.cudnn.deterministic = True @torch.cuda.amp.custom_fwd(cast_inputs=torch.float32) def normalize_keypoints( kpts: torch.Tensor, size: Optional[torch.Tensor] = None ) -> torch.Tensor: if size is None: size = 1 + kpts.max(-2).values - kpts.min(-2).values elif not isinstance(size, torch.Tensor): size = torch.tensor(size, device=kpts.device, dtype=kpts.dtype) size = size.to(kpts) shift = size / 2 scale = size.max(-1).values / 2 kpts = (kpts - shift[..., None, :]) / scale[..., None, None] return kpts def rotate_half(x: torch.Tensor) -> torch.Tensor: x = x.unflatten(-1, (-1, 2)) x1, x2 = x.unbind(dim=-1) return torch.stack((-x2, x1), dim=-1).flatten(start_dim=-2) def apply_cached_rotary_emb(freqs: torch.Tensor, t: torch.Tensor) -> torch.Tensor: return (t * freqs[0]) + (rotate_half(t) * freqs[1]) class LearnableFourierPositionalEncoding(nn.Module): def __init__(self, M: int, dim: int, F_dim: int = None, gamma: float = 1.0) -> None: super().__init__() F_dim = F_dim if F_dim is not None else dim self.gamma = gamma self.Wr = nn.Linear(M, F_dim // 2, bias=False) nn.init.normal_(self.Wr.weight.data, mean=0, std=self.gamma**-2) def forward(self, x: torch.Tensor) -> torch.Tensor: """encode position vector""" projected = self.Wr(x) cosines, sines = torch.cos(projected), torch.sin(projected) emb = torch.stack([cosines, sines], 0).unsqueeze(-3) return emb.repeat_interleave(2, dim=-1) class TokenConfidence(nn.Module): def __init__(self, dim: int) -> None: super().__init__() self.token = nn.Sequential(nn.Linear(dim, 1), nn.Sigmoid()) self.loss_fn = nn.BCEWithLogitsLoss(reduction="none") def forward(self, desc0: torch.Tensor, desc1: torch.Tensor): """get confidence tokens""" return ( self.token(desc0.detach()).squeeze(-1), self.token(desc1.detach()).squeeze(-1), ) def loss(self, desc0, desc1, la_now, la_final): logit0 = self.token[0](desc0.detach()).squeeze(-1) logit1 = self.token[0](desc1.detach()).squeeze(-1) la_now, la_final = la_now.detach(), la_final.detach() correct0 = ( la_final[:, :-1, :].max(-1).indices == la_now[:, :-1, :].max(-1).indices ) correct1 = ( la_final[:, :, :-1].max(-2).indices == la_now[:, :, :-1].max(-2).indices ) return ( self.loss_fn(logit0, correct0.float()).mean(-1) + self.loss_fn(logit1, correct1.float()).mean(-1) ) / 2.0 class Attention(nn.Module): def __init__(self, allow_flash: bool) -> None: super().__init__() if allow_flash and not FLASH_AVAILABLE: warnings.warn( "FlashAttention is not available. For optimal speed, " "consider installing torch >= 2.0 or flash-attn.", stacklevel=2, ) self.enable_flash = allow_flash and FLASH_AVAILABLE if FLASH_AVAILABLE: torch.backends.cuda.enable_flash_sdp(allow_flash) def forward(self, q, k, v, mask: Optional[torch.Tensor] = None) -> torch.Tensor: if self.enable_flash and q.device.type == "cuda": # use torch 2.0 scaled_dot_product_attention with flash if FLASH_AVAILABLE: args = [x.half().contiguous() for x in [q, k, v]] v = F.scaled_dot_product_attention(*args, attn_mask=mask).to(q.dtype) return v if mask is None else v.nan_to_num() elif FLASH_AVAILABLE: args = [x.contiguous() for x in [q, k, v]] v = F.scaled_dot_product_attention(*args, attn_mask=mask) return v if mask is None else v.nan_to_num() else: s = q.shape[-1] ** -0.5 sim = torch.einsum("...id,...jd->...ij", q, k) * s if mask is not None: sim.masked_fill(~mask, -float("inf")) attn = F.softmax(sim, -1) return torch.einsum("...ij,...jd->...id", attn, v) class SelfBlock(nn.Module): def __init__( self, embed_dim: int, num_heads: int, flash: bool = False, bias: bool = True ) -> None: super().__init__() self.embed_dim = embed_dim self.num_heads = num_heads assert self.embed_dim % num_heads == 0 self.head_dim = self.embed_dim // num_heads self.Wqkv = nn.Linear(embed_dim, 3 * embed_dim, bias=bias) self.inner_attn = Attention(flash) self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias) self.ffn = nn.Sequential( nn.Linear(2 * embed_dim, 2 * embed_dim), nn.LayerNorm(2 * embed_dim, elementwise_affine=True), nn.GELU(), nn.Linear(2 * embed_dim, embed_dim), ) def forward( self, x: torch.Tensor, encoding: torch.Tensor, mask: Optional[torch.Tensor] = None, ) -> torch.Tensor: qkv = self.Wqkv(x) qkv = qkv.unflatten(-1, (self.num_heads, -1, 3)).transpose(1, 2) q, k, v = qkv[..., 0], qkv[..., 1], qkv[..., 2] q = apply_cached_rotary_emb(encoding, q) k = apply_cached_rotary_emb(encoding, k) context = self.inner_attn(q, k, v, mask=mask) message = self.out_proj(context.transpose(1, 2).flatten(start_dim=-2)) return x + self.ffn(torch.cat([x, message], -1)) class CrossBlock(nn.Module): def __init__( self, embed_dim: int, num_heads: int, flash: bool = False, bias: bool = True ) -> None: super().__init__() self.heads = num_heads dim_head = embed_dim // num_heads self.scale = dim_head**-0.5 inner_dim = dim_head * num_heads self.to_qk = nn.Linear(embed_dim, inner_dim, bias=bias) self.to_v = nn.Linear(embed_dim, inner_dim, bias=bias) self.to_out = nn.Linear(inner_dim, embed_dim, bias=bias) self.ffn = nn.Sequential( nn.Linear(2 * embed_dim, 2 * embed_dim), nn.LayerNorm(2 * embed_dim, elementwise_affine=True), nn.GELU(), nn.Linear(2 * embed_dim, embed_dim), ) if flash and FLASH_AVAILABLE: self.flash = Attention(True) else: self.flash = None def map_(self, func: Callable, x0: torch.Tensor, x1: torch.Tensor): return func(x0), func(x1) def forward( self, x0: torch.Tensor, x1: torch.Tensor, mask: Optional[torch.Tensor] = None ) -> List[torch.Tensor]: qk0, qk1 = self.map_(self.to_qk, x0, x1) v0, v1 = self.map_(self.to_v, x0, x1) qk0, qk1, v0, v1 = map( lambda t: t.unflatten(-1, (self.heads, -1)).transpose(1, 2), (qk0, qk1, v0, v1), ) if self.flash is not None and qk0.device.type == "cuda": m0 = self.flash(qk0, qk1, v1, mask) m1 = self.flash( qk1, qk0, v0, mask.transpose(-1, -2) if mask is not None else None ) else: qk0, qk1 = qk0 * self.scale**0.5, qk1 * self.scale**0.5 sim = torch.einsum("bhid, bhjd -> bhij", qk0, qk1) if mask is not None: sim = sim.masked_fill(~mask, -float("inf")) attn01 = F.softmax(sim, dim=-1) attn10 = F.softmax(sim.transpose(-2, -1).contiguous(), dim=-1) m0 = torch.einsum("bhij, bhjd -> bhid", attn01, v1) m1 = torch.einsum("bhji, bhjd -> bhid", attn10.transpose(-2, -1), v0) if mask is not None: m0, m1 = m0.nan_to_num(), m1.nan_to_num() m0, m1 = self.map_(lambda t: t.transpose(1, 2).flatten(start_dim=-2), m0, m1) m0, m1 = self.map_(self.to_out, m0, m1) x0 = x0 + self.ffn(torch.cat([x0, m0], -1)) x1 = x1 + self.ffn(torch.cat([x1, m1], -1)) return x0, x1 class TransformerLayer(nn.Module): def __init__(self, *args, **kwargs): super().__init__() self.self_attn = SelfBlock(*args, **kwargs) self.cross_attn = CrossBlock(*args, **kwargs) def forward( self, desc0, desc1, encoding0, encoding1, mask0: Optional[torch.Tensor] = None, mask1: Optional[torch.Tensor] = None, ): if mask0 is not None and mask1 is not None: return self.masked_forward(desc0, desc1, encoding0, encoding1, mask0, mask1) else: desc0 = self.self_attn(desc0, encoding0) desc1 = self.self_attn(desc1, encoding1) return self.cross_attn(desc0, desc1) # This part is compiled and allows padding inputs def masked_forward(self, desc0, desc1, encoding0, encoding1, mask0, mask1): mask = mask0 & mask1.transpose(-1, -2) mask0 = mask0 & mask0.transpose(-1, -2) mask1 = mask1 & mask1.transpose(-1, -2) desc0 = self.self_attn(desc0, encoding0, mask0) desc1 = self.self_attn(desc1, encoding1, mask1) return self.cross_attn(desc0, desc1, mask) def sigmoid_log_double_softmax( sim: torch.Tensor, z0: torch.Tensor, z1: torch.Tensor ) -> torch.Tensor: """create the log assignment matrix from logits and similarity""" b, m, n = sim.shape certainties = F.logsigmoid(z0) + F.logsigmoid(z1).transpose(1, 2) scores0 = F.log_softmax(sim, 2) scores1 = F.log_softmax(sim.transpose(-1, -2).contiguous(), 2).transpose(-1, -2) scores = sim.new_full((b, m + 1, n + 1), 0) scores[:, :m, :n] = scores0 + scores1 + certainties scores[:, :-1, -1] = F.logsigmoid(-z0.squeeze(-1)) scores[:, -1, :-1] = F.logsigmoid(-z1.squeeze(-1)) return scores class MatchAssignment(nn.Module): def __init__(self, dim: int) -> None: super().__init__() self.dim = dim self.matchability = nn.Linear(dim, 1, bias=True) self.final_proj = nn.Linear(dim, dim, bias=True) def forward(self, desc0: torch.Tensor, desc1: torch.Tensor): """build assignment matrix from descriptors""" mdesc0, mdesc1 = self.final_proj(desc0), self.final_proj(desc1) _, _, d = mdesc0.shape mdesc0, mdesc1 = mdesc0 / d**0.25, mdesc1 / d**0.25 sim = torch.einsum("bmd,bnd->bmn", mdesc0, mdesc1) z0 = self.matchability(desc0) z1 = self.matchability(desc1) scores = sigmoid_log_double_softmax(sim, z0, z1) return scores, sim def get_matchability(self, desc: torch.Tensor): return torch.sigmoid(self.matchability(desc)).squeeze(-1) def filter_matches(scores: torch.Tensor, th: float): """obtain matches from a log assignment matrix [Bx M+1 x N+1]""" max0, max1 = scores[:, :-1, :-1].max(2), scores[:, :-1, :-1].max(1) m0, m1 = max0.indices, max1.indices indices0 = torch.arange(m0.shape[1], device=m0.device)[None] indices1 = torch.arange(m1.shape[1], device=m1.device)[None] mutual0 = indices0 == m1.gather(1, m0) mutual1 = indices1 == m0.gather(1, m1) max0_exp = max0.values.exp() zero = max0_exp.new_tensor(0) mscores0 = torch.where(mutual0, max0_exp, zero) mscores1 = torch.where(mutual1, mscores0.gather(1, m1), zero) valid0 = mutual0 & (mscores0 > th) valid1 = mutual1 & valid0.gather(1, m1) m0 = torch.where(valid0, m0, -1) m1 = torch.where(valid1, m1, -1) return m0, m1, mscores0, mscores1 class LightGlue(nn.Module): default_conf = { "name": "lightglue", # just for interfacing "input_dim": 256, # input descriptor dimension (autoselected from weights) "add_scale_ori": False, "descriptor_dim": 256, "n_layers": 9, "num_heads": 4, "flash": False, # enable FlashAttention if available. "mp": False, # enable mixed precision "depth_confidence": -1, # early stopping, disable with -1 "width_confidence": -1, # point pruning, disable with -1 "filter_threshold": 0.0, # match threshold "checkpointed": False, "weights": "superpoint_lightglue", # either a path or the name of pretrained weights (disk, ...) "weights_from_version": "v0.1_arxiv", "loss": { "gamma": 1.0, "fn": "nll", "nll_balancing": 0.5, }, } required_data_keys = ["keypoints0", "keypoints1", "descriptors0", "descriptors1"] url = "https://github.com/cvg/LightGlue/releases/download/{}/{}.pth" def __init__(self, conf) -> None: super().__init__() self.conf = conf = OmegaConf.merge(self.default_conf, conf) if conf.input_dim != conf.descriptor_dim: self.input_proj = nn.Linear(conf.input_dim, conf.descriptor_dim, bias=True) else: self.input_proj = nn.Identity() head_dim = conf.descriptor_dim // conf.num_heads self.posenc = LearnableFourierPositionalEncoding( 2 + 2 * conf.add_scale_ori, head_dim, head_dim ) h, n, d = conf.num_heads, conf.n_layers, conf.descriptor_dim self.transformers = nn.ModuleList( [TransformerLayer(d, h, conf.flash) for _ in range(n)] ) self.log_assignment = nn.ModuleList([MatchAssignment(d) for _ in range(n)]) self.token_confidence = nn.ModuleList( [TokenConfidence(d) for _ in range(n - 1)] ) self.loss_fn = NLLLoss(conf.loss) state_dict = None if conf.weights is not None: # weights can be either a path or an existing file from official LG if Path(conf.weights).exists(): state_dict = torch.load(conf.weights, map_location="cpu") elif (Path(DATA_PATH) / conf.weights).exists(): state_dict = torch.load( str(DATA_PATH / conf.weights), map_location="cpu" ) elif (Path('weights') / (conf.weights + '.pth')).exists(): state_dict = torch.load( str(Path('weights') / (conf.weights + '.pth')), map_location="cpu" ) print(f"Readed weights from {Path('weights') / (conf.weights + '.pth')}") else: fname = ( f"{conf.weights}_{conf.weights_from_version}".replace(".", "-") + ".pth" ) state_dict = torch.hub.load_state_dict_from_url( self.url.format(conf.weights_from_version, conf.weights), file_name=fname, ) if state_dict: # rename old state dict entries for i in range(self.conf.n_layers): pattern = f"self_attn.{i}", f"transformers.{i}.self_attn" state_dict = {k.replace(*pattern): v for k, v in state_dict.items()} pattern = f"cross_attn.{i}", f"transformers.{i}.cross_attn" state_dict = {k.replace(*pattern): v for k, v in state_dict.items()} self.load_state_dict(state_dict, strict=False) print(f"Loaded weights from {conf.weights}") def compile(self, mode="reduce-overhead"): if self.conf.width_confidence != -1: warnings.warn( "Point pruning is partially disabled for compiled forward.", stacklevel=2, ) for i in range(self.conf.n_layers): self.transformers[i] = torch.compile( self.transformers[i], mode=mode, fullgraph=True ) def forward(self, data: dict) -> dict: for key in self.required_data_keys: assert key in data, f"Missing key {key} in data" kpts0, kpts1 = data["keypoints0"], data["keypoints1"] b, m, _ = kpts0.shape b, n, _ = kpts1.shape device = kpts0.device # if "view0" in data.keys() and "view1" in data.keys(): size0 = data["resize0"][:, [1, 0]] size1 = data["resize1"][:, [1, 0]] kpts0 = normalize_keypoints(kpts0, size0).clone() kpts1 = normalize_keypoints(kpts1, size1).clone() if self.conf.add_scale_ori: sc0, o0 = data["scales0"], data["oris0"] sc1, o1 = data["scales1"], data["oris1"] kpts0 = torch.cat( [ kpts0, sc0 if sc0.dim() == 3 else sc0[..., None], o0 if o0.dim() == 3 else o0[..., None], ], -1, ) kpts1 = torch.cat( [ kpts1, sc1 if sc1.dim() == 3 else sc1[..., None], o1 if o1.dim() == 3 else o1[..., None], ], -1, ) desc0 = data["descriptors0"].contiguous() desc1 = data["descriptors1"].contiguous() assert desc0.shape[-1] == self.conf.input_dim assert desc1.shape[-1] == self.conf.input_dim if torch.is_autocast_enabled(): desc0 = desc0.half() desc1 = desc1.half() desc0 = self.input_proj(desc0) desc1 = self.input_proj(desc1) # cache positional embeddings encoding0 = self.posenc(kpts0) encoding1 = self.posenc(kpts1) # GNN + final_proj + assignment do_early_stop = self.conf.depth_confidence > 0 and not self.training do_point_pruning = self.conf.width_confidence > 0 and not self.training all_desc0, all_desc1 = [], [] if do_point_pruning: ind0 = torch.arange(0, m, device=device)[None] ind1 = torch.arange(0, n, device=device)[None] # We store the index of the layer at which pruning is detected. prune0 = torch.ones_like(ind0) prune1 = torch.ones_like(ind1) token0, token1 = None, None for i in range(self.conf.n_layers): if self.conf.checkpointed and self.training: desc0, desc1 = checkpoint( self.transformers[i], desc0, desc1, encoding0, encoding1 ) else: desc0, desc1 = self.transformers[i](desc0, desc1, encoding0, encoding1) if self.training or i == self.conf.n_layers - 1: all_desc0.append(desc0) all_desc1.append(desc1) continue # no early stopping or adaptive width at last layer # only for eval if do_early_stop: assert b == 1 token0, token1 = self.token_confidence[i](desc0, desc1) if self.check_if_stop(token0[..., :m, :], token1[..., :n, :], i, m + n): break if do_point_pruning: assert b == 1 scores0 = self.log_assignment[i].get_matchability(desc0) prunemask0 = self.get_pruning_mask(token0, scores0, i) keep0 = torch.where(prunemask0)[1] ind0 = ind0.index_select(1, keep0) desc0 = desc0.index_select(1, keep0) encoding0 = encoding0.index_select(-2, keep0) prune0[:, ind0] += 1 scores1 = self.log_assignment[i].get_matchability(desc1) prunemask1 = self.get_pruning_mask(token1, scores1, i) keep1 = torch.where(prunemask1)[1] ind1 = ind1.index_select(1, keep1) desc1 = desc1.index_select(1, keep1) encoding1 = encoding1.index_select(-2, keep1) prune1[:, ind1] += 1 desc0, desc1 = desc0[..., :m, :], desc1[..., :n, :] scores, _ = self.log_assignment[i](desc0, desc1) m0, m1, mscores0, mscores1 = filter_matches(scores, self.conf.filter_threshold) matches, mscores = [], [] for k in range(b): if self.training: break valid = m0[k] > -1 m_indices_0 = torch.where(valid)[0] m_indices_1 = m0[k][valid] if do_point_pruning: m_indices_0 = ind0[k, m_indices_0] m_indices_1 = ind1[k, m_indices_1] matches.append(torch.stack([m_indices_0, m_indices_1], -1)) mscores.append(mscores0[k][valid]) if do_point_pruning: m0_ = torch.full((b, m), -1, device=m0.device, dtype=m0.dtype) m1_ = torch.full((b, n), -1, device=m1.device, dtype=m1.dtype) m0_[:, ind0] = torch.where(m0 == -1, -1, ind1.gather(1, m0.clamp(min=0))) m1_[:, ind1] = torch.where(m1 == -1, -1, ind0.gather(1, m1.clamp(min=0))) mscores0_ = torch.zeros((b, m), device=mscores0.device) mscores1_ = torch.zeros((b, n), device=mscores1.device) mscores0_[:, ind0] = mscores0 mscores1_[:, ind1] = mscores1 m0, m1, mscores0, mscores1 = m0_, m1_, mscores0_, mscores1_ else: prune0 = torch.ones_like(mscores0) * self.conf.n_layers prune1 = torch.ones_like(mscores1) * self.conf.n_layers pred = { "matches0": m0, "matches1": m1, "matching_scores0": mscores0, "matching_scores1": mscores1, "ref_descriptors0": torch.stack(all_desc0, 1), "ref_descriptors1": torch.stack(all_desc1, 1), "log_assignment": scores, "stop": i + 1, "matches": matches, "scores": mscores, "prune0": prune0, "prune1": prune1, } return pred def confidence_threshold(self, layer_index: int) -> float: """scaled confidence threshold""" threshold = 0.8 + 0.1 * np.exp(-4.0 * layer_index / self.conf.n_layers) return np.clip(threshold, 0, 1) def get_pruning_mask( self, confidences: torch.Tensor, scores: torch.Tensor, layer_index: int ) -> torch.Tensor: """mask points which should be removed""" keep = scores > (1 - self.conf.width_confidence) if confidences is not None: # Low-confidence points are never pruned. keep |= confidences <= self.confidence_thresholds[layer_index] return keep def check_if_stop( self, confidences0: torch.Tensor, confidences1: torch.Tensor, layer_index: int, num_points: int, ) -> torch.Tensor: """evaluate stopping condition""" confidences = torch.cat([confidences0, confidences1], -1) threshold = self.confidence_thresholds[layer_index] ratio_confident = 1.0 - (confidences < threshold).float().sum() / num_points return ratio_confident > self.conf.depth_confidence def pruning_min_kpts(self, device: torch.device): if self.conf.flash and FLASH_AVAILABLE and device.type == "cuda": return self.pruning_keypoint_thresholds["flash"] else: return self.pruning_keypoint_thresholds[device.type] def loss(self, pred, data): def loss_params(pred, i): la, _ = self.log_assignment[i]( pred["ref_descriptors0"][:, i], pred["ref_descriptors1"][:, i] ) return { "log_assignment": la, } sum_weights = 1.0 nll, gt_weights, loss_metrics = self.loss_fn(loss_params(pred, -1), data) N = pred["ref_descriptors0"].shape[1] losses = {"total": nll, "last": nll.clone().detach(), **loss_metrics} if self.training: losses["confidence"] = 0.0 # B = pred['log_assignment'].shape[0] losses["row_norm"] = pred["log_assignment"].exp()[:, :-1].sum(2).mean(1) for i in range(N - 1): params_i = loss_params(pred, i) nll, _, _ = self.loss_fn(params_i, data, weights=gt_weights) if self.conf.loss.gamma > 0.0: weight = self.conf.loss.gamma ** (N - i - 1) else: weight = i + 1 sum_weights += weight losses["total"] = losses["total"] + nll * weight losses["confidence"] += self.token_confidence[i].loss( pred["ref_descriptors0"][:, i], pred["ref_descriptors1"][:, i], params_i["log_assignment"], pred["log_assignment"], ) / (N - 1) del params_i losses["total"] /= sum_weights # confidences if self.training: losses["total"] = losses["total"] + losses["confidence"] if not self.training: # add metrics metrics = matcher_metrics(pred, data) else: metrics = {} return losses, metrics __main_model__ = LightGlue