projects/deta/modeling/deformable_transformer.py

# coding=utf-8
# Copyright 2022 The IDEA Authors. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import math
import torch
import torch.nn as nn

from detectron2.layers.nms import batched_nms

from detrex.layers import (
    FFN,
    BaseTransformerLayer,
    MultiheadAttention,
    MultiScaleDeformableAttention,
    TransformerLayerSequence,
    box_cxcywh_to_xyxy
)
from detrex.utils import inverse_sigmoid


class DeformableDetrTransformerEncoder(TransformerLayerSequence):
    def __init__(
        self,
        embed_dim: int = 256,
        num_heads: int = 8,
        feedforward_dim: int = 1024,
        attn_dropout: float = 0.1,
        ffn_dropout: float = 0.1,
        num_layers: int = 6,
        post_norm: bool = False,
        num_feature_levels: int = 4,
    ):
        super(DeformableDetrTransformerEncoder, self).__init__(
            transformer_layers=BaseTransformerLayer(
                attn=MultiScaleDeformableAttention(
                    embed_dim=embed_dim,
                    num_heads=num_heads,
                    dropout=attn_dropout,
                    batch_first=True,
                    num_levels=num_feature_levels,
                ),
                ffn=FFN(
                    embed_dim=embed_dim,
                    feedforward_dim=feedforward_dim,
                    output_dim=embed_dim,
                    num_fcs=2,
                    ffn_drop=ffn_dropout,
                ),
                norm=nn.LayerNorm(embed_dim),
                operation_order=("self_attn", "norm", "ffn", "norm"),
            ),
            num_layers=num_layers,
        )
        self.embed_dim = self.layers[0].embed_dim
        self.pre_norm = self.layers[0].pre_norm

        if post_norm:
            self.post_norm_layer = nn.LayerNorm(self.embed_dim)
        else:
            self.post_norm_layer = None

    def forward(
        self,
        query,
        key,
        value,
        query_pos=None,
        key_pos=None,
        attn_masks=None,
        query_key_padding_mask=None,
        key_padding_mask=None,
        **kwargs,
    ):

        for layer in self.layers:
            query = layer(
                query,
                key,
                value,
                query_pos=query_pos,
                attn_masks=attn_masks,
                query_key_padding_mask=query_key_padding_mask,
                key_padding_mask=key_padding_mask,
                **kwargs,
            )

        if self.post_norm_layer is not None:
            query = self.post_norm_layer(query)
        return query


class DeformableDetrTransformerDecoder(TransformerLayerSequence):
    def __init__(
        self,
        embed_dim: int = 256,
        num_heads: int = 8,
        feedforward_dim: int = 1024,
        attn_dropout: float = 0.1,
        ffn_dropout: float = 0.1,
        num_layers: int = 6,
        return_intermediate: bool = True,
        num_feature_levels: int = 4,
    ):
        super(DeformableDetrTransformerDecoder, self).__init__(
            transformer_layers=BaseTransformerLayer(
                attn=[
                    MultiheadAttention(
                        embed_dim=embed_dim,
                        num_heads=num_heads,
                        attn_drop=attn_dropout,
                        batch_first=True,
                    ),
                    MultiScaleDeformableAttention(
                        embed_dim=embed_dim,
                        num_heads=num_heads,
                        dropout=attn_dropout,
                        batch_first=True,
                        num_levels=num_feature_levels,
                    ),
                ],
                ffn=FFN(
                    embed_dim=embed_dim,
                    feedforward_dim=feedforward_dim,
                    output_dim=embed_dim,
                    ffn_drop=ffn_dropout,
                ),
                norm=nn.LayerNorm(embed_dim),
                operation_order=("self_attn", "norm", "cross_attn", "norm", "ffn", "norm"),
            ),
            num_layers=num_layers,
        )
        self.return_intermediate = return_intermediate

        self.bbox_embed = None
        self.class_embed = None

    def forward(
        self,
        query,
        key,
        value,
        query_pos=None,
        key_pos=None,
        attn_masks=None,
        query_key_padding_mask=None,
        key_padding_mask=None,
        reference_points=None,
        valid_ratios=None,
        **kwargs,
    ):
        output = query

        intermediate = []
        intermediate_reference_points = []
        for layer_idx, layer in enumerate(self.layers):
            if reference_points.shape[-1] == 4:
                reference_points_input = (
                    reference_points[:, :, None]
                    * torch.cat([valid_ratios, valid_ratios], -1)[:, None]
                )
            else:
                assert reference_points.shape[-1] == 2
                reference_points_input = reference_points[:, :, None] * valid_ratios[:, None]

            output = layer(
                output,
                key,
                value,
                query_pos=query_pos,
                key_pos=key_pos,
                attn_masks=attn_masks,
                query_key_padding_mask=query_key_padding_mask,
                key_padding_mask=key_padding_mask,
                reference_points=reference_points_input,
                **kwargs,
            )

            if self.bbox_embed is not None:
                tmp = self.bbox_embed[layer_idx](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


class DeformableDetrTransformer(nn.Module):
    """Transformer module for Deformable DETR

    Args:
        encoder (nn.Module): encoder module.
        decoder (nn.Module): decoder module.
        as_two_stage (bool): whether to use two-stage transformer. Default False.
        num_feature_levels (int): number of feature levels. Default 4.
        two_stage_num_proposals (int): number of proposals in two-stage transformer. Default 300.
            Only used when as_two_stage is True.
    """

    def __init__(
        self,
        encoder=None,
        decoder=None,
        num_feature_levels=4,
        as_two_stage=False,
        two_stage_num_proposals=300,
        assign_first_stage=True,
    ):
        super(DeformableDetrTransformer, self).__init__()
        self.encoder = encoder
        self.decoder = decoder
        self.num_feature_levels = num_feature_levels
        self.as_two_stage = as_two_stage
        self.two_stage_num_proposals = two_stage_num_proposals

        # DETA implementation
        self.assign_first_stage = assign_first_stage

        self.embed_dim = self.encoder.embed_dim

        self.level_embeds = nn.Parameter(torch.Tensor(self.num_feature_levels, self.embed_dim))

        if self.as_two_stage:
            self.enc_output = nn.Linear(self.embed_dim, self.embed_dim)
            self.enc_output_norm = nn.LayerNorm(self.embed_dim)
            self.pos_trans = nn.Linear(self.embed_dim * 2, self.embed_dim * 2)
            self.pos_trans_norm = nn.LayerNorm(self.embed_dim * 2)
            # DETA implementation
            self.pix_trans = nn.Linear(self.embed_dim, self.embed_dim)
            self.pix_trans_norm = nn.LayerNorm(self.embed_dim)
        else:
            self.reference_points = nn.Linear(self.embed_dim, 2)

        self.init_weights()

    def init_weights(self):
        for p in self.parameters():
            if p.dim() > 1:
                nn.init.xavier_uniform_(p)
        for m in self.modules():
            if isinstance(m, MultiScaleDeformableAttention):
                m.init_weights()
        if not self.as_two_stage:
            nn.init.xavier_normal_(self.reference_points.weight.data, gain=1.0)
            nn.init.constant_(self.reference_points.bias.data, 0.0)
        nn.init.normal_(self.level_embeds)

    def gen_encoder_output_proposals(self, memory, memory_padding_mask, spatial_shapes):
        N, S, C = memory.shape
        proposals = []
        _cur = 0
        level_ids = []
        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
            level_ids.append(grid.new_ones(H * W, dtype=torch.long) * lvl)

        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))
        level_ids = torch.cat(level_ids)
        return output_memory, output_proposals, level_ids

    @staticmethod
    def get_reference_points(spatial_shapes, valid_ratios, device):
        """Get the reference points used in decoder.

        Args:
            spatial_shapes (Tensor): The shape of all
                feature maps, has shape (num_level, 2).
            valid_ratios (Tensor): The ratios of valid
                points on the feature map, has shape
                (bs, num_levels, 2)
            device (obj:`device`): The device where
                reference_points should be.

        Returns:
            Tensor: reference points used in decoder, has \
                shape (bs, num_keys, num_levels, 2).
        """
        reference_points_list = []
        for lvl, (H, W) in enumerate(spatial_shapes):
            #  TODO  check this 0.5
            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 get_valid_ratio(self, mask):
        """Get the valid ratios of feature maps of all levels."""
        _, 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 get_proposal_pos_embed(self, proposals, num_pos_feats=128, temperature=10000):
        """Get the position embedding of proposal."""
        scale = 2 * math.pi
        dim_t = torch.arange(num_pos_feats, dtype=torch.float32, device=proposals.device)
        dim_t = temperature ** (2 * torch.div(dim_t, 2, rounding_mode="floor") / 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 forward(
        self,
        multi_level_feats,
        multi_level_masks,
        multi_level_pos_embeds,
        query_embed,
        **kwargs,
    ):
        assert self.as_two_stage or query_embed is not None

        feat_flatten = []
        mask_flatten = []
        lvl_pos_embed_flatten = []
        spatial_shapes = []
        for lvl, (feat, mask, pos_embed) in enumerate(
            zip(multi_level_feats, multi_level_masks, multi_level_pos_embeds)
        ):
            bs, c, h, w = feat.shape
            spatial_shape = (h, w)
            spatial_shapes.append(spatial_shape)

            feat = feat.flatten(2).transpose(1, 2)  # bs, hw, c
            mask = mask.flatten(1)
            pos_embed = pos_embed.flatten(2).transpose(1, 2)  # bs, hw, c
            lvl_pos_embed = pos_embed + self.level_embeds[lvl].view(1, 1, -1)
            lvl_pos_embed_flatten.append(lvl_pos_embed)
            feat_flatten.append(feat)
            mask_flatten.append(mask)
        feat_flatten = torch.cat(feat_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=feat_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 multi_level_masks], 1)

        reference_points = self.get_reference_points(
            spatial_shapes, valid_ratios, device=feat.device
        )

        memory = self.encoder(
            query=feat_flatten,
            key=None,
            value=None,
            query_pos=lvl_pos_embed_flatten,
            query_key_padding_mask=mask_flatten,
            spatial_shapes=spatial_shapes,
            reference_points=reference_points,
            level_start_index=level_start_index,
            valid_ratios=valid_ratios,
            **kwargs,
        )

        bs, _, c = memory.shape
        if self.as_two_stage:
            output_memory, output_proposals, level_ids = self.gen_encoder_output_proposals(
                memory, mask_flatten, spatial_shapes
            )

            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
            proposal_logit = enc_outputs_class[..., 0]

            if self.assign_first_stage:
                proposal_boxes = box_cxcywh_to_xyxy(enc_outputs_coord_unact.sigmoid().float()).clamp(0, 1)
                topk_proposals = []
                for b in range(bs):
                    prop_boxes_b = proposal_boxes[b]
                    prop_logits_b = proposal_logit[b]

                    # pre-nms per-level topk
                    pre_nms_topk = 1000
                    pre_nms_inds = []
                    for lvl in range(len(spatial_shapes)):
                        lvl_mask = level_ids == lvl
                        pre_nms_inds.append(torch.topk(prop_logits_b.sigmoid() * lvl_mask, pre_nms_topk)[1])
                    pre_nms_inds = torch.cat(pre_nms_inds)

                    # nms on topk indices
                    post_nms_inds = batched_nms(prop_boxes_b[pre_nms_inds], prop_logits_b[pre_nms_inds], level_ids[pre_nms_inds], 0.9)
                    keep_inds = pre_nms_inds[post_nms_inds]

                    if len(keep_inds) < self.two_stage_num_proposals:
                        print(f'[WARNING] nms proposals ({len(keep_inds)}) < {self.two_stage_num_proposals}, running naive topk')
                        keep_inds = torch.topk(proposal_logit[b], topk)[1]

                    # keep top Q/L indices for L levels
                    q_per_l = topk // len(spatial_shapes)
                    is_level_ordered = level_ids[keep_inds][None] == torch.arange(len(spatial_shapes), device=level_ids.device)[:,None]  # LS
                    keep_inds_mask = is_level_ordered & (is_level_ordered.cumsum(1) <= q_per_l)  # LS
                    keep_inds_mask = keep_inds_mask.any(0)  # S

                    # pad to Q indices (might let ones filtered from pre-nms sneak by... unlikely because we pick high conf anyways)
                    if keep_inds_mask.sum() < topk:
                        num_to_add = topk - keep_inds_mask.sum()
                        pad_inds = (~keep_inds_mask).nonzero()[:num_to_add]
                        keep_inds_mask[pad_inds] = True

                    # index
                    keep_inds_topk = keep_inds[keep_inds_mask]
                    topk_proposals.append(keep_inds_topk)
                topk_proposals = torch.stack(topk_proposals)
            else:
                topk_proposals = torch.topk(proposal_logit, 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_pos, query = torch.split(pos_trans_out, c, dim=2)

            topk_feats = torch.stack([output_memory[b][topk_proposals[b]] for b in range(bs)]).detach()
            query = query + self.pix_trans_norm(self.pix_trans(topk_feats))

        else:
            query_pos, query = torch.split(query_embed, c, dim=1)
            query_pos = query_pos.unsqueeze(0).expand(bs, -1, -1)
            query = query.unsqueeze(0).expand(bs, -1, -1)
            reference_points = self.reference_points(query_pos).sigmoid()
            init_reference_out = reference_points

        # decoder
        inter_states, inter_references = self.decoder(
            query=query,  # bs, num_queries, embed_dims
            key=None,  # bs, num_tokens, embed_dims
            value=memory,  # bs, num_tokens, embed_dims
            query_pos=query_pos,
            key_padding_mask=mask_flatten,  # bs, num_tokens
            reference_points=reference_points,  # num_queries, 4
            spatial_shapes=spatial_shapes,  # nlvl, 2
            level_start_index=level_start_index,  # nlvl
            valid_ratios=valid_ratios,  # bs, nlvl, 2
            **kwargs,
        )

        inter_references_out = inter_references
        if self.as_two_stage:
            return (
                inter_states,
                init_reference_out,
                inter_references_out,
                enc_outputs_class,
                enc_outputs_coord_unact,
                output_proposals.sigmoid(),
            )
        return inter_states, init_reference_out, inter_references_out, None, None, None