MMDetection源码解析：Faster RCNN(3)--RPN Head类
Faster RCNN配置文件faster_rcnn_r50_fpn.py中的
backbone=dict( type='ResNet', depth=50, num_stages=4, out_indices=(0, 1, 2, 3), frozen_stages=1, norm_cfg=dict(type='BN', requires_grad=True), norm_eval=True, style='pytorch'),
设置了Backbone为ResNet.
neck=dict( type='FPN', in_channels=[256, 512, 1024, 2048], out_channels=256, num_outs=5),
设置了Neck为FPN.Backbone和Neck比较简单,就不详细介绍了,详细介绍一下RPN Head.
rpn_head=dict( type='RPNHead', in_channels=256, feat_channels=256, anchor_generator=dict( type='AnchorGenerator', scales=[8], ratios=[0.5, 1.0, 2.0], strides=[4, 8, 16, 32, 64]), bbox_coder=dict( type='DeltaXYWHBBoxCoder', target_means=[.0, .0, .0, .0], target_stds=[1.0, 1.0, 1.0, 1.0]), loss_cls=dict( type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0), loss_bbox=dict(type='L1Loss', loss_weight=1.0)),
以上文件设置了RPN Head为RPNHead(RPNHead定义在mmdetection/mmdet/models/dense_heads/rpn_head.py文件里面);同时定义了anchor_generator为AnchorGenerator(AnchorGenerator定义在mmdetection/mmdet/core/anchor/anchor_generator.py里面),指定了Anchor生成的方式;bbox_coder为DeltaXYWHBBoxCoder(DeltaXYWHBBoxCoder定义在mmdetection/mmdet/core/bbox/coder/delta_xywh_bbox_coder.py里面),指定了BBox的编码解码方式;loss_cls和loss_bbox设置了BBox的分类损失和回归损失.
rpn_head.py文件内容如下：
import torch import torch.nn as nn import torch.nn.functional as F from mmcv.cnn import normal_init from mmcv.ops import batched_nms from ..builder import HEADS from .anchor_head import AnchorHead from .rpn_test_mixin import RPNTestMixin @HEADS.register_module() class RPNHead(RPNTestMixin, AnchorHead): """RPN head. Args: in_channels (int): Number of channels in the input feature map. """ # noqa: W605 def __init__(self, in_channels, **kwargs): super(RPNHead, self).__init__(1, in_channels, **kwargs) def _init_layers(self): """Initialize layers of the head.""" self.rpn_conv = nn.Conv2d( self.in_channels, self.feat_channels, 3, padding=1) self.rpn_cls = nn.Conv2d(self.feat_channels, self.num_anchors * self.cls_out_channels, 1) self.rpn_reg = nn.Conv2d(self.feat_channels, self.num_anchors * 4, 1) def init_weights(self): """Initialize weights of the head.""" normal_init(self.rpn_conv, std=0.01) normal_init(self.rpn_cls, std=0.01) normal_init(self.rpn_reg, std=0.01) def forward_single(self, x): """Forward feature map of a single scale level.""" x = self.rpn_conv(x) x = F.relu(x, inplace=True) rpn_cls_score = self.rpn_cls(x) rpn_bbox_pred = self.rpn_reg(x) return rpn_cls_score, rpn_bbox_pred def loss(self, cls_scores, bbox_preds, gt_bboxes, img_metas, gt_bboxes_ignore=None): """Compute losses of the head. Args: cls_scores (list[Tensor]): Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W) bbox_preds (list[Tensor]): Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W) gt_bboxes (list[Tensor]): Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format. img_metas (list[dict]): Meta information of each image, e.g., image size, scaling factor, etc. gt_bboxes_ignore (None | list[Tensor]): specify which bounding boxes can be ignored when computing the loss. Returns: dict[str, Tensor]: A dictionary of loss components. """ losses = super(RPNHead, self).loss( cls_scores, bbox_preds, gt_bboxes, None, img_metas, gt_bboxes_ignore=gt_bboxes_ignore) return dict( loss_rpn_cls=losses['loss_cls'], loss_rpn_bbox=losses['loss_bbox']) def _get_bboxes_single(self, cls_scores, bbox_preds, mlvl_anchors, img_shape, scale_factor, cfg, rescale=False): """Transform outputs for a single batch item into bbox predictions. Args: cls_scores (list[Tensor]): Box scores for each scale level Has shape (num_anchors * num_classes, H, W). bbox_preds (list[Tensor]): Box energies / deltas for each scale level with shape (num_anchors * 4, H, W). mlvl_anchors (list[Tensor]): Box reference for each scale level with shape (num_total_anchors, 4). img_shape (tuple[int]): Shape of the input image, (height, width, 3). scale_factor (ndarray): Scale factor of the image arange as (w_scale, h_scale, w_scale, h_scale). cfg (mmcv.Config): Test / postprocessing configuration, if None, test_cfg would be used. rescale (bool): If True, return boxes in original image space. Returns: Tensor: Labeled boxes in shape (n, 5), where the first 4 columns are bounding box positions (tl_x, tl_y, br_x, br_y) and the 5-th column is a score between 0 and 1. """ cfg = self.test_cfg if cfg is None else cfg # bboxes from different level should be independent during NMS, # level_ids are used as labels for batched NMS to separate them level_ids = [] mlvl_scores = [] mlvl_bbox_preds = [] mlvl_valid_anchors = [] for idx in range(len(cls_scores)): rpn_cls_score = cls_scores[idx] rpn_bbox_pred = bbox_preds[idx] assert rpn_cls_score.size()[-2:] == rpn_bbox_pred.size()[-2:] rpn_cls_score = rpn_cls_score.permute(1, 2, 0) if self.use_sigmoid_cls: rpn_cls_score = rpn_cls_score.reshape(-1) scores = rpn_cls_score.sigmoid() else: rpn_cls_score = rpn_cls_score.reshape(-1, 2) # We set FG labels to [0, num_class-1] and BG label to # num_class in RPN head since mmdet v2.5, which is unified to # be consistent with other head since mmdet v2.0. In mmdet v2.0 # to v2.4 we keep BG label as 0 and FG label as 1 in rpn head. scores = rpn_cls_score.softmax(dim=1)[:, 0] rpn_bbox_pred = rpn_bbox_pred.permute(1, 2, 0).reshape(-1, 4) anchors = mlvl_anchors[idx] if cfg.nms_pre > 0 and scores.shape[0] > cfg.nms_pre: # sort is faster than topk # _, topk_inds = scores.topk(cfg.nms_pre) ranked_scores, rank_inds = scores.sort(descending=True) topk_inds = rank_inds[:cfg.nms_pre] scores = ranked_scores[:cfg.nms_pre] rpn_bbox_pred = rpn_bbox_pred[topk_inds, :] anchors = anchors[topk_inds, :] mlvl_scores.append(scores) mlvl_bbox_preds.append(rpn_bbox_pred) mlvl_valid_anchors.append(anchors) level_ids.append( scores.new_full((scores.size(0), ), idx, dtype=torch.long)) scores = torch.cat(mlvl_scores) anchors = torch.cat(mlvl_valid_anchors) rpn_bbox_pred = torch.cat(mlvl_bbox_preds) proposals = self.bbox_coder.decode( anchors, rpn_bbox_pred, max_shape=img_shape) ids = torch.cat(level_ids) if cfg.min_bbox_size > 0: w = proposals[:, 2] - proposals[:, 0] h = proposals[:, 3] - proposals[:, 1] valid_inds = torch.nonzero( (w >= cfg.min_bbox_size) & (h >= cfg.min_bbox_size), as_tuple=False).squeeze() if valid_inds.sum().item() != len(proposals): proposals = proposals[valid_inds, :] scores = scores[valid_inds] ids = ids[valid_inds] # TODO: remove the hard coded nms type nms_cfg = dict(type='nms', iou_threshold=cfg.nms_thr) dets, keep = batched_nms(proposals, scores, ids, nms_cfg) return dets[:cfg.nms_post]
RPNHead的主要包括以下函数：
(1)__init__(),初始化函数.
(2)_init_layers(),该函数主要实现Head的Layer的初始化,从代码可知,有3个Layer：卷积层rpn_conv,分类层rpn_cls,回归层rpn_reg.
这3层都用到了nn.Conv2d这个类(定义在torch/nn/modules/conv.py里面),默认的第1个参数in_channels,第2个参数是out_channels,第3个参数是kernel_size.rpn_cls的out_channels是每一个点的Anchor数量乘以cls_out_channels(等于类的数量+1).rpn_reg的out_channels是Anchor的数量乘以4,因为一个BBox由4个输出决定.
(3)init_weights(),初始化权重.
(4)forward_single(),把FPN的1个特征图Forward,也就是输出这一层的分类和BBox预测.
(5)loss(),计算损失函数.
(6)_get_bboxes_single(),根据forward_single()的输出,以及每一层的Anchor,得到最终的预测结果,再进行NMS.
