# ------------------------------------------------------------------------ # Copyright (c) 2022 megvii-research. All Rights Reserved. # ------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ import copy from typing import Optional, List import math import torch import torch.nn.functional as F from torch import nn, Tensor from torch.nn.init import xavier_uniform_, constant_, uniform_, normal_ from models.structures import Boxes, matched_boxlist_iou, pairwise_iou from util.misc import inverse_sigmoid from util.box_ops import box_cxcywh_to_xyxy from models.ops.modules import MSDeformAttn class DeformableTransformer(nn.Module): def __init__(self, d_model=256, nhead=8, num_encoder_layers=6, num_decoder_layers=6, dim_feedforward=1024, dropout=0.1, activation="relu", return_intermediate_dec=False, num_feature_levels=4, dec_n_points=4, enc_n_points=4, two_stage=False, two_stage_num_proposals=300, decoder_self_cross=True, sigmoid_attn=False, extra_track_attn=False, memory_bank=False): super().__init__() self.new_frame_adaptor = None self.d_model = d_model self.nhead = nhead self.two_stage = two_stage self.two_stage_num_proposals = two_stage_num_proposals encoder_layer = DeformableTransformerEncoderLayer(d_model, dim_feedforward, dropout, activation, num_feature_levels, nhead, enc_n_points, sigmoid_attn=sigmoid_attn) self.encoder = DeformableTransformerEncoder(encoder_layer, num_encoder_layers) decoder_layer = DeformableTransformerDecoderLayer(d_model, dim_feedforward, dropout, activation, num_feature_levels, nhead, dec_n_points, decoder_self_cross, sigmoid_attn=sigmoid_attn, extra_track_attn=extra_track_attn, memory_bank=memory_bank) self.decoder = DeformableTransformerDecoder(decoder_layer, num_decoder_layers, return_intermediate_dec) self.level_embed = nn.Parameter(torch.Tensor(num_feature_levels, d_model)) if two_stage: self.enc_output = nn.Linear(d_model, d_model) self.enc_output_norm = nn.LayerNorm(d_model) self.pos_trans = nn.Linear(d_model * 2, d_model * 2) self.pos_trans_norm = nn.LayerNorm(d_model * 2) self._reset_parameters() def _reset_parameters(self): for p in self.parameters(): if p.dim() > 1: nn.init.xavier_uniform_(p) for m in self.modules(): if isinstance(m, MSDeformAttn): m._reset_parameters() normal_(self.level_embed) def get_proposal_pos_embed(self, proposals): num_pos_feats = 128 temperature = 10000 scale = 2 * math.pi dim_t = torch.arange(num_pos_feats, dtype=torch.float32, device=proposals.device) dim_t = temperature ** (2 * (dim_t // 2) / num_pos_feats) # N, L, 4 proposals = proposals.sigmoid() * scale # N, L, 4, 128 pos = proposals[:, :, :, None] / dim_t # N, L, 4, 64, 2 pos = torch.stack((pos[:, :, :, 0::2].sin(), pos[:, :, :, 1::2].cos()), dim=4).flatten(2) return pos def gen_encoder_output_proposals(self, memory, memory_padding_mask, spatial_shapes): N_, S_, C_ = memory.shape base_scale = 4.0 proposals = [] _cur = 0 for lvl, (H_, W_) in enumerate(spatial_shapes): mask_flatten_ = memory_padding_mask[:, _cur:(_cur + H_ * W_)].view(N_, H_, W_, 1) valid_H = torch.sum(~mask_flatten_[:, :, 0, 0], 1) valid_W = torch.sum(~mask_flatten_[:, 0, :, 0], 1) grid_y, grid_x = torch.meshgrid(torch.linspace(0, H_ - 1, H_, dtype=torch.float32, device=memory.device), torch.linspace(0, W_ - 1, W_, dtype=torch.float32, device=memory.device)) grid = torch.cat([grid_x.unsqueeze(-1), grid_y.unsqueeze(-1)], -1) scale = torch.cat([valid_W.unsqueeze(-1), valid_H.unsqueeze(-1)], 1).view(N_, 1, 1, 2) grid = (grid.unsqueeze(0).expand(N_, -1, -1, -1) + 0.5) / scale wh = torch.ones_like(grid) * 0.05 * (2.0 ** lvl) proposal = torch.cat((grid, wh), -1).view(N_, -1, 4) proposals.append(proposal) _cur += (H_ * W_) output_proposals = torch.cat(proposals, 1) output_proposals_valid = ((output_proposals > 0.01) & (output_proposals < 0.99)).all(-1, keepdim=True) output_proposals = torch.log(output_proposals / (1 - output_proposals)) output_proposals = output_proposals.masked_fill(memory_padding_mask.unsqueeze(-1), float('inf')) output_proposals = output_proposals.masked_fill(~output_proposals_valid, float('inf')) output_memory = memory output_memory = output_memory.masked_fill(memory_padding_mask.unsqueeze(-1), float(0)) output_memory = output_memory.masked_fill(~output_proposals_valid, float(0)) output_memory = self.enc_output_norm(self.enc_output(output_memory)) return output_memory, output_proposals def get_valid_ratio(self, mask): _, H, W = mask.shape valid_H = torch.sum(~mask[:, :, 0], 1) valid_W = torch.sum(~mask[:, 0, :], 1) valid_ratio_h = valid_H.float() / H valid_ratio_w = valid_W.float() / W valid_ratio = torch.stack([valid_ratio_w, valid_ratio_h], -1) return valid_ratio def forward(self, srcs, masks, pos_embeds, query_embed=None, ref_pts=None, mem_bank=None, mem_bank_pad_mask=None, attn_mask=None): assert self.two_stage or query_embed is not None # prepare input for encoder src_flatten = [] mask_flatten = [] lvl_pos_embed_flatten = [] spatial_shapes = [] for lvl, (src, mask, pos_embed) in enumerate(zip(srcs, masks, pos_embeds)): bs, c, h, w = src.shape spatial_shape = (h, w) spatial_shapes.append(spatial_shape) src = src.flatten(2).transpose(1, 2) mask = mask.flatten(1) pos_embed = pos_embed.flatten(2).transpose(1, 2) lvl_pos_embed = pos_embed + self.level_embed[lvl].view(1, 1, -1) lvl_pos_embed_flatten.append(lvl_pos_embed) src_flatten.append(src) mask_flatten.append(mask) src_flatten = torch.cat(src_flatten, 1) mask_flatten = torch.cat(mask_flatten, 1) lvl_pos_embed_flatten = torch.cat(lvl_pos_embed_flatten, 1) spatial_shapes = torch.as_tensor(spatial_shapes, dtype=torch.long, device=src_flatten.device) level_start_index = torch.cat((spatial_shapes.new_zeros((1, )), spatial_shapes.prod(1).cumsum(0)[:-1])) valid_ratios = torch.stack([self.get_valid_ratio(m) for m in masks], 1) # encoder memory = self.encoder(src_flatten, spatial_shapes, level_start_index, valid_ratios, lvl_pos_embed_flatten, mask_flatten) # prepare input for decoder bs, _, c = memory.shape if self.two_stage: output_memory, output_proposals = self.gen_encoder_output_proposals(memory, mask_flatten, spatial_shapes) # hack implementation for two-stage Deformable DETR enc_outputs_class = self.decoder.class_embed[self.decoder.num_layers](output_memory) enc_outputs_coord_unact = self.decoder.bbox_embed[self.decoder.num_layers](output_memory) + output_proposals topk = self.two_stage_num_proposals topk_proposals = torch.topk(enc_outputs_class[..., 0], topk, dim=1)[1] topk_coords_unact = torch.gather(enc_outputs_coord_unact, 1, topk_proposals.unsqueeze(-1).repeat(1, 1, 4)) topk_coords_unact = topk_coords_unact.detach() reference_points = topk_coords_unact.sigmoid() init_reference_out = reference_points pos_trans_out = self.pos_trans_norm(self.pos_trans(self.get_proposal_pos_embed(topk_coords_unact))) query_embed, tgt = torch.split(pos_trans_out, c, dim=2) else: tgt = query_embed.unsqueeze(0).expand(bs, -1, -1) reference_points = ref_pts.unsqueeze(0).expand(bs, -1, -1) init_reference_out = reference_points # decoder hs, inter_references = self.decoder(tgt, reference_points, memory, spatial_shapes, level_start_index, valid_ratios, mask_flatten, mem_bank, mem_bank_pad_mask, attn_mask) inter_references_out = inter_references if self.two_stage: return hs, init_reference_out, inter_references_out, enc_outputs_class, enc_outputs_coord_unact return hs, init_reference_out, inter_references_out, None, None class DeformableTransformerEncoderLayer(nn.Module): def __init__(self, d_model=256, d_ffn=1024, dropout=0.1, activation="relu", n_levels=4, n_heads=8, n_points=4, sigmoid_attn=False): super().__init__() # self attention self.self_attn = MSDeformAttn(d_model, n_levels, n_heads, n_points, sigmoid_attn=sigmoid_attn) self.dropout1 = nn.Dropout(dropout) self.norm1 = nn.LayerNorm(d_model) # ffn self.linear1 = nn.Linear(d_model, d_ffn) self.activation = _get_activation_fn(activation) self.dropout_relu = ReLUDropout(dropout, True) self.linear2 = nn.Linear(d_ffn, d_model) self.dropout3 = nn.Dropout(dropout) self.norm2 = nn.LayerNorm(d_model) @staticmethod def with_pos_embed(tensor, pos): return tensor if pos is None else tensor + pos def forward_ffn(self, src): src2 = self.linear2(self.dropout_relu(self.linear1(src))) src = src + self.dropout3(src2) src = self.norm2(src) return src def forward(self, src, pos, reference_points, spatial_shapes, level_start_index, padding_mask=None): # self attention src2 = self.self_attn(self.with_pos_embed(src, pos), reference_points, src, spatial_shapes, level_start_index, padding_mask) src = src + self.dropout1(src2) src = self.norm1(src) # ffn src = self.forward_ffn(src) return src class DeformableTransformerEncoder(nn.Module): def __init__(self, encoder_layer, num_layers): super().__init__() self.layers = _get_clones(encoder_layer, num_layers) self.num_layers = num_layers @staticmethod def get_reference_points(spatial_shapes, valid_ratios, device): reference_points_list = [] for lvl, (H_, W_) in enumerate(spatial_shapes): ref_y, ref_x = torch.meshgrid(torch.linspace(0.5, H_ - 0.5, H_, dtype=torch.float32, device=device), torch.linspace(0.5, W_ - 0.5, W_, dtype=torch.float32, device=device)) ref_y = ref_y.reshape(-1)[None] / (valid_ratios[:, None, lvl, 1] * H_) ref_x = ref_x.reshape(-1)[None] / (valid_ratios[:, None, lvl, 0] * W_) ref = torch.stack((ref_x, ref_y), -1) reference_points_list.append(ref) reference_points = torch.cat(reference_points_list, 1) reference_points = reference_points[:, :, None] * valid_ratios[:, None] return reference_points def forward(self, src, spatial_shapes, level_start_index, valid_ratios, pos=None, padding_mask=None): output = src reference_points = self.get_reference_points(spatial_shapes, valid_ratios, device=src.device) for _, layer in enumerate(self.layers): output = layer(output, pos, reference_points, spatial_shapes, level_start_index, padding_mask) return output class ReLUDropout(torch.nn.Dropout): def forward(self, input): return relu_dropout(input, p=self.p, training=self.training, inplace=self.inplace) def relu_dropout(x, p=0, inplace=False, training=False): if not training or p == 0: return x.clamp_(min=0) if inplace else x.clamp(min=0) mask = (x < 0) | (torch.rand_like(x) > 1 - p) return x.masked_fill_(mask, 0).div_(1 - p) if inplace else x.masked_fill(mask, 0).div(1 - p) class DeformableTransformerDecoderLayer(nn.Module): def __init__(self, d_model=256, d_ffn=1024, dropout=0.1, activation="relu", n_levels=4, n_heads=8, n_points=4, self_cross=True, sigmoid_attn=False, extra_track_attn=False, memory_bank=False): super().__init__() self.self_cross = self_cross self.num_head = n_heads self.memory_bank = memory_bank # cross attention self.cross_attn = MSDeformAttn(d_model, n_levels, n_heads, n_points, sigmoid_attn=sigmoid_attn) self.dropout1 = nn.Dropout(dropout) self.norm1 = nn.LayerNorm(d_model) self.self_attn = nn.MultiheadAttention(d_model, n_heads, dropout=dropout) self.dropout2 = nn.Dropout(dropout) self.norm2 = nn.LayerNorm(d_model) # ffn self.linear1 = nn.Linear(d_model, d_ffn) self.activation = _get_activation_fn(activation) self.dropout_relu = ReLUDropout(dropout, True) self.linear2 = nn.Linear(d_ffn, d_model) self.dropout4 = nn.Dropout(dropout) self.norm3 = nn.LayerNorm(d_model) # memory bank if self.memory_bank: self.temporal_attn = nn.MultiheadAttention(d_model, 8, dropout=0) self.temporal_fc1 = nn.Linear(d_model, d_ffn) self.temporal_fc2 = nn.Linear(d_ffn, d_model) self.temporal_norm1 = nn.LayerNorm(d_model) self.temporal_norm2 = nn.LayerNorm(d_model) position = torch.arange(5).unsqueeze(1) div_term = torch.exp(torch.arange(0, d_model, 2) * (-math.log(10000.0) / d_model)) pe = torch.zeros(5, 1, d_model) pe[:, 0, 0::2] = torch.sin(position * div_term) pe[:, 0, 1::2] = torch.cos(position * div_term) self.register_buffer('pe', pe) # update track query_embed self.extra_track_attn = extra_track_attn if self.extra_track_attn: print('Training with Extra Self Attention in Every Decoder.', flush=True) self.update_attn = nn.MultiheadAttention(d_model, n_heads, dropout=dropout) self.dropout5 = nn.Dropout(dropout) self.norm4 = nn.LayerNorm(d_model) if self_cross: print('Training with Self-Cross Attention.') else: print('Training with Cross-Self Attention.') @staticmethod def with_pos_embed(tensor, pos): return tensor if pos is None else tensor + pos def forward_ffn(self, tgt): tgt2 = self.linear2(self.dropout_relu(self.linear1(tgt))) tgt = tgt + self.dropout4(tgt2) tgt = self.norm3(tgt) return tgt def _forward_self_attn(self, tgt, query_pos, attn_mask=None): q = k = self.with_pos_embed(tgt, query_pos) if attn_mask is not None: tgt2 = self.self_attn(q.transpose(0, 1), k.transpose(0, 1), tgt.transpose(0, 1), attn_mask=attn_mask)[0].transpose(0, 1) else: tgt2 = self.self_attn(q.transpose(0, 1), k.transpose(0, 1), tgt.transpose(0, 1))[0].transpose(0, 1) tgt = tgt + self.dropout2(tgt2) return self.norm2(tgt) def _forward_self_cross(self, tgt, query_pos, reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask=None, attn_mask=None): # self attention tgt = self._forward_self_attn(tgt, query_pos, attn_mask) # cross attention tgt2 = self.cross_attn(self.with_pos_embed(tgt, query_pos), reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask) tgt = tgt + self.dropout1(tgt2) tgt = self.norm1(tgt) # ffn tgt = self.forward_ffn(tgt) return tgt def _forward_cross_self(self, tgt, query_pos, reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask=None, attn_mask=None): # cross attention tgt2 = self.cross_attn(self.with_pos_embed(tgt, query_pos), reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask) tgt = tgt + self.dropout1(tgt2) tgt = self.norm1(tgt) # self attention tgt = self._forward_self_attn(tgt, query_pos, attn_mask) # ffn tgt = self.forward_ffn(tgt) return tgt def forward(self, tgt, query_pos, reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask=None, mem_bank=None, mem_bank_pad_mask=None, attn_mask=None): if self.self_cross: return self._forward_self_cross(tgt, query_pos, reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask, attn_mask) return self._forward_cross_self(tgt, query_pos, reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask, attn_mask) def pos2posemb(pos, num_pos_feats=64, temperature=10000): scale = 2 * math.pi pos = pos * scale dim_t = torch.arange(num_pos_feats, dtype=torch.float32, device=pos.device) dim_t = temperature ** (2 * (dim_t // 2) / num_pos_feats) posemb = pos[..., None] / dim_t posemb = torch.stack((posemb[..., 0::2].sin(), posemb[..., 1::2].cos()), dim=-1).flatten(-3) return posemb class DeformableTransformerDecoder(nn.Module): def __init__(self, decoder_layer, num_layers, return_intermediate=False): super().__init__() self.layers = _get_clones(decoder_layer, num_layers) self.num_layers = num_layers self.return_intermediate = return_intermediate # hack implementation for iterative bounding box refinement and two-stage Deformable DETR self.bbox_embed = None self.class_embed = None def forward(self, tgt, reference_points, src, src_spatial_shapes, src_level_start_index, src_valid_ratios, src_padding_mask=None, mem_bank=None, mem_bank_pad_mask=None, attn_mask=None): output = tgt intermediate = [] intermediate_reference_points = [] for lid, layer in enumerate(self.layers): if reference_points.shape[-1] == 4: reference_points_input = reference_points[:, :, None] \ * torch.cat([src_valid_ratios, src_valid_ratios], -1)[:, None] else: assert reference_points.shape[-1] == 2 reference_points_input = reference_points[:, :, None] * src_valid_ratios[:, None] query_pos = pos2posemb(reference_points) output = layer(output, query_pos, reference_points_input, src, src_spatial_shapes, src_level_start_index, src_padding_mask, mem_bank, mem_bank_pad_mask, attn_mask) # hack implementation for iterative bounding box refinement if self.bbox_embed is not None: tmp = self.bbox_embed[lid](output) if reference_points.shape[-1] == 4: new_reference_points = tmp + inverse_sigmoid(reference_points) new_reference_points = new_reference_points.sigmoid() else: assert reference_points.shape[-1] == 2 new_reference_points = tmp new_reference_points[..., :2] = tmp[..., :2] + inverse_sigmoid(reference_points) new_reference_points = new_reference_points.sigmoid() reference_points = new_reference_points.detach() if self.return_intermediate: intermediate.append(output) intermediate_reference_points.append(reference_points) if self.return_intermediate: return torch.stack(intermediate), torch.stack(intermediate_reference_points) return output, reference_points def _get_clones(module, N): return nn.ModuleList([copy.deepcopy(module) for i in range(N)]) def _get_activation_fn(activation): """Return an activation function given a string""" if activation == "relu": return nn.ReLU(True) if activation == "gelu": return F.gelu if activation == "glu": return F.glu raise RuntimeError(F"activation should be relu/gelu, not {activation}.") def build_deforamble_transformer(args): return DeformableTransformer( d_model=args.hidden_dim, nhead=args.nheads, num_encoder_layers=args.enc_layers, num_decoder_layers=args.dec_layers, dim_feedforward=args.dim_feedforward, dropout=args.dropout, activation="relu", return_intermediate_dec=True, num_feature_levels=args.num_feature_levels, dec_n_points=args.dec_n_points, enc_n_points=args.enc_n_points, two_stage=args.two_stage, two_stage_num_proposals=args.num_queries, decoder_self_cross=not args.decoder_cross_self, sigmoid_attn=args.sigmoid_attn, extra_track_attn=args.extra_track_attn, memory_bank=args.memory_bank_type == 'MemoryBankFeat' )