Faster RCNN 官方源碼解讀
導(dǎo)讀
?本文詳細的介紹了 torchvision 中的 FasterRCNN 代碼實現(xiàn)，并分析了作者認為重要的知識點，GeneralizedRCNN的代碼以及FasterRCNN的訓(xùn)練等。幫助入門的小伙伴更好的理解模型細節(jié)的問題。
目前 pytorch 已經(jīng)在 torchvision 模塊集成了 FasterRCNN 和 MaskRCNN 代碼。考慮到幫助各位小伙伴理解模型細節(jié)問題，本文分析一下 FasterRCNN 代碼，幫助新手理解 Two-Stage 檢測中的主要問題。
這篇文章默認讀者已經(jīng)對 FasterRCNN 原理有一定了解。否則請先點擊閱讀上一篇文章：
torchvision 中 FasterRCNN 代碼文檔如下：
https://pytorch.org/docs/stable/torchvision/models.html#faster-r-cnnpytorch.org
在 python 中裝好 torchvision 后，輸入以下命令即可查看版本和代碼位置：
import torchvision
print(torchvision.__version__)
# '0.6.0'
print(torchvision.__path__)
# ['/usr/local/lib/python3.7/site-packages/torchvision']代碼結(jié)構(gòu)
圖1
作為 torchvision 中目標(biāo)檢測基類，GeneralizedRCNN 繼承了 torch.nn.Module，后續(xù) FasterRCNN 、MaskRCNN 都繼承 GeneralizedRCNN。
GeneralizedRCNN
GeneralizedRCNN 繼承基類 nn.Module 。首先來看看基類 GeneralizedRCNN 的代碼：
class GeneralizedRCNN(nn.Module):
def __init__(self, backbone, rpn, roi_heads, transform):
super(GeneralizedRCNN, self).__init__()
self.transform = transform
self.backbone = backbone
self.rpn = rpn
self.roi_heads = roi_heads
# used only on torchscript mode
self._has_warned = False
@torch.jit.unused
def eager_outputs(self, losses, detections):
# type: (Dict[str, Tensor], List[Dict[str, Tensor]]) -> Tuple[Dict[str, Tensor], List[Dict[str, Tensor]]]
if self.training:
return losses
return detections
def forward(self, images, targets=None):
if self.training and targets is None:
raise ValueError("In training mode, targets should be passed")
original_image_sizes = torch.jit.annotate(List[Tuple[int, int]], [])
for img in images:
val = img.shape[-2:]
assert len(val) == 2
original_image_sizes.append((val[0], val[1]))
images, targets = self.transform(images, targets)
features = self.backbone(images.tensors)
if isinstance(features, torch.Tensor):
features = OrderedDict([('0', features)])
proposals, proposal_losses = self.rpn(images, features, targets)
detections, detector_losses = self.roi_heads(features, proposals, images.image_sizes, targets)
detections = self.transform.postprocess(detections, images.image_sizes, original_image_sizes)
losses = {}
losses.update(detector_losses)
losses.update(proposal_losses)
if torch.jit.is_scripting():
if not self._has_warned:
warnings.warn("RCNN always returns a (Losses, Detections) tuple in scripting")
self._has_warned = True
return (losses, detections)
else:
return self.eager_outputs(losses, detections)對于 GeneralizedRCNN 類，其中有4個重要的接口：
transform backbone rpn roi_heads
transform
# GeneralizedRCNN.forward(...)
for img in images:
val = img.shape[-2:]
assert len(val) == 2
original_image_sizes.append((val[0], val[1]))
images, targets = self.transform(images, targets)
圖2 transform接口
transform主要做2件事：
將輸入進行標(biāo)準(zhǔn)化（如FasterRCNN是對 輸入減 image_mean 再除 image_std） 將圖像縮放到固定大小（同時也要對應(yīng)縮放 targets 中標(biāo)記框 ）
需要說明，由于把縮放后的圖像輸入網(wǎng)絡(luò)，那么網(wǎng)絡(luò)輸出的檢測框也是在縮放后的圖像上的。但是實際中我們需要的是在原始圖像的檢測框，為了對應(yīng)起來，所以需要記錄變換前original_images_sizes 。
圖3
這里解釋一下為何要縮放圖像。對于 FasterRCNN，從純理論上來說確實可以支持任意大小的圖片。但是實際中，如果輸入圖像太大（如6000x4000）會直接撐爆內(nèi)存。考慮到工程問題，縮放是一個比較穩(wěn)妥的折衷選擇。
backbone + rpn + roi_heads
圖4
完成圖像縮放之后其實才算是正式進入網(wǎng)絡(luò)流程。接下來有4個步驟：
將 transform 后的圖像輸入到 backbone 模塊提取特征圖
# GeneralizedRCNN.forward(...)
features = self.backbone(images.tensors)
backbone 一般為 VGG、ResNet、MobileNet 等網(wǎng)絡(luò)。
然后經(jīng)過 rpn 模塊生成 proposals 和 proposal_losses
# GeneralizedRCNN.forward(...)
features = self.backbone(images.tensors)
接著進入 roi_heads 模塊（即 roi_pooling + 分類）
# GeneralizedRCNN.forward(...)
detections, detector_losses =
self.roi_heads(features, proposals, images.image_sizes, targets
最后經(jīng) postprocess 模塊（進行 NMS，同時將 box 通過 original_images_size映射回原圖）
# GeneralizedRCNN.forward(...)
detections = self.transform.postprocess(detections, images.image_sizes, original_image_sizes)
FasterRCNN
FasterRCNN 繼承基類 GeneralizedRCNN。
class FasterRCNN(GeneralizedRCNN):
def __init__(self, backbone, num_classes=None,
# transform parameters
min_size=800, max_size=1333,
image_mean=None, image_std=None,
# RPN parameters
rpn_anchor_generator=None, rpn_head=None,
rpn_pre_nms_top_n_train=2000, rpn_pre_nms_top_n_test=1000,
rpn_post_nms_top_n_train=2000, rpn_post_nms_top_n_test=1000,
rpn_nms_thresh=0.7,
rpn_fg_iou_thresh=0.7, rpn_bg_iou_thresh=0.3,
rpn_batch_size_per_image=256, rpn_positive_fraction=0.5,
# Box parameters
box_roi_pool=None, box_head=None, box_predictor=None,
box_score_thresh=0.05, box_nms_thresh=0.5, box_detections_per_img=100,
box_fg_iou_thresh=0.5, box_bg_iou_thresh=0.5,
box_batch_size_per_image=512, box_positive_fraction=0.25,
bbox_reg_weights=None):
out_channels = backbone.out_channels
if rpn_anchor_generator is None:
anchor_sizes = ((32,), (64,), (128,), (256,), (512,))
aspect_ratios = ((0.5, 1.0, 2.0),) * len(anchor_sizes)
rpn_anchor_generator = AnchorGenerator(
anchor_sizes, aspect_ratios
)
if rpn_head is None:
rpn_head = RPNHead(
out_channels, rpn_anchor_generator.num_anchors_per_location()[0]
)
rpn_pre_nms_top_n = dict(training=rpn_pre_nms_top_n_train, testing=rpn_pre_nms_top_n_test)
rpn_post_nms_top_n = dict(training=rpn_post_nms_top_n_train, testing=rpn_post_nms_top_n_test)
rpn = RegionProposalNetwork(
rpn_anchor_generator, rpn_head,
rpn_fg_iou_thresh, rpn_bg_iou_thresh,
rpn_batch_size_per_image, rpn_positive_fraction,
rpn_pre_nms_top_n, rpn_post_nms_top_n, rpn_nms_thresh)
if box_roi_pool is None:
box_roi_pool = MultiScaleRoIAlign(
featmap_names=['0', '1', '2', '3'],
output_size=7,
sampling_ratio=2)
if box_head is None:
resolution = box_roi_pool.output_size[0]
representation_size = 1024
box_head = TwoMLPHead(
out_channels * resolution ** 2,
representation_size)
if box_predictor is None:
representation_size = 1024
box_predictor = FastRCNNPredictor(
representation_size,
num_classes)
roi_heads = RoIHeads(
# Box
box_roi_pool, box_head, box_predictor,
box_fg_iou_thresh, box_bg_iou_thresh,
box_batch_size_per_image, box_positive_fraction,
bbox_reg_weights,
box_score_thresh, box_nms_thresh, box_detections_per_img)
if image_mean is None:
image_mean = [0.485, 0.456, 0.406]
if image_std is None:
image_std = [0.229, 0.224, 0.225]
transform = GeneralizedRCNNTransform(min_size, max_size, image_mean, image_std)
super(FasterRCNN, self).__init__(backbone, rpn, roi_heads, transform)FasterRCNN 實現(xiàn)了 GeneralizedRCNN 中的 transform、backbone、rpn、roi_heads 接口：
# FasterRCNN.__init__(...)
super(FasterRCNN, self).__init__(backbone, rpn, roi_heads, transform)
對于 transform 接口，使用 GeneralizedRCNNTransform 實現(xiàn)。從代碼變量名可以明顯看到包含：
與縮放相關(guān)參數(shù)：min_size + max_size 與歸一化相關(guān)參數(shù)：image_mean + image_std（對輸入[0, 1]減去image_mean再除以image_std）
# FasterRCNN.__init__(...)
if image_mean is None:
image_mean = [0.485, 0.456, 0.406]
if image_std is None:
image_std = [0.229, 0.224, 0.225]
transform = GeneralizedRCNNTransform(min_size, max_size, image_mean, image_std)對于 backbone 使用 ResNet50 + FPN 結(jié)構(gòu)：
def fasterrcnn_resnet50_fpn(pretrained=False, progress=True, num_classes=91, pretrained_backbone=True, **kwargs):
if pretrained:
# no need to download the backbone if pretrained is set
pretrained_backbone = False
backbone = resnet_fpn_backbone('resnet50', pretrained_backbone)
model = FasterRCNN(backbone, num_classes, **kwargs)
if pretrained:
state_dict = load_state_dict_from_url(model_urls['fasterrcnn_resnet50_fpn_coco'], progress=progress)
model.load_state_dict(state_dict)
return modelResNet: Deep Residual Learning for Image Recognition
FPN: Feature Pyramid Networks for Object Detection
圖5 FPN
接下來重點介紹 rpn 接口的實現(xiàn)。首先是 rpn_anchor_generator :
# FasterRCNN.__init__(...)
if rpn_anchor_generator is None:
anchor_sizes = ((32,), (64,), (128,), (256,), (512,))
aspect_ratios = ((0.5, 1.0, 2.0),) * len(anchor_sizes)
rpn_anchor_generator = AnchorGenerator(
anchor_sizes, aspect_ratios
)
對于普通的 FasterRCNN 只需要將 feature_map 輸入到 rpn 網(wǎng)絡(luò)生成 proposals 即可。但是由于加入 FPN，需要將多個 feature_map 逐個輸入到 rpn 網(wǎng)絡(luò)。
圖6
接下來看看 AnchorGenerator 具體實現(xiàn)：
class AnchorGenerator(nn.Module):
......
def generate_anchors(self, scales, aspect_ratios, dtype=torch.float32, device="cpu"):
# type: (List[int], List[float], int, Device) # noqa: F821
scales = torch.as_tensor(scales, dtype=dtype, device=device)
aspect_ratios = torch.as_tensor(aspect_ratios, dtype=dtype, device=device)
h_ratios = torch.sqrt(aspect_ratios)
w_ratios = 1 / h_ratios
ws = (w_ratios[:, None] * scales[None, :]).view(-1)
hs = (h_ratios[:, None] * scales[None, :]).view(-1)
base_anchors = torch.stack([-ws, -hs, ws, hs], dim=1) / 2
return base_anchors.round()
def set_cell_anchors(self, dtype, device):
# type: (int, Device) -> None # noqa: F821
......
cell_anchors = [
self.generate_anchors(
sizes,
aspect_ratios,
dtype,
device
)
for sizes, aspect_ratios in zip(self.sizes, self.aspect_ratios)
]
self.cell_anchors = cell_anchors
首先，每個位置有 5 種 anchor_size 和 3 種 aspect_ratios，所以每個位置生成 15 個 base_anchors：
[ -23., -11., 23., 11.]
[ -16., -16., 16., 16.] # w = h = 32, ratio = 1
[ -11., -23., 11., 23.]
[ -45., -23., 45., 23.]
[ -32., -32., 32., 32.] # w = h = 64, ratio = 1
[ -23., -45., 23., 45.]
[ -91., -45., 91., 45.]
[ -64., -64., 64., 64.] # w = h = 128, ratio = 1
[ -45., -91., 45., 91.]
[-181., -91., 181., 91.]
[-128., -128., 128., 128.] # w = h = 256, ratio = 1
[ -91., -181., 91., 181.]
[-362., -181., 362., 181.]
[-256., -256., 256., 256.] # w = h = 512, ratio = 1
[-181., -362., 181., 362.]
注意 base_anchors 的中心都是 點，如下圖所示：
圖7 base_anchor（此圖只畫了32/64/128的base_anchor）
接著來看 AnchorGenerator.grid_anchors 函數(shù)：
# AnchorGenerator
def grid_anchors(self, grid_sizes, strides):
# type: (List[List[int]], List[List[Tensor]])
anchors = []
cell_anchors = self.cell_anchors
assert cell_anchors is not None
for size, stride, base_anchors in zip(
grid_sizes, strides, cell_anchors
):
grid_height, grid_width = size
stride_height, stride_width = stride
device = base_anchors.device
# For output anchor, compute [x_center, y_center, x_center, y_center]
shifts_x = torch.arange(
0, grid_width, dtype=torch.float32, device=device
) * stride_width
shifts_y = torch.arange(
0, grid_height, dtype=torch.float32, device=device
) * stride_height
shift_y, shift_x = torch.meshgrid(shifts_y, shifts_x)
shift_x = shift_x.reshape(-1)
shift_y = shift_y.reshape(-1)
shifts = torch.stack((shift_x, shift_y, shift_x, shift_y), dim=1)
# For every (base anchor, output anchor) pair,
# offset each zero-centered base anchor by the center of the output anchor.
anchors.append(
(shifts.view(-1, 1, 4) + base_anchors.view(1, -1, 4)).reshape(-1, 4)
)
return anchors
def forward(self, image_list, feature_maps):
# type: (ImageList, List[Tensor])
grid_sizes = list([feature_map.shape[-2:] for feature_map in feature_maps])
image_size = image_list.tensors.shape[-2:]
dtype, device = feature_maps[0].dtype, feature_maps[0].device
strides = [[torch.tensor(image_size[0] / g[0], dtype=torch.int64, device=device),
torch.tensor(image_size[1] / g[1], dtype=torch.int64, device=device)] for g in grid_sizes]
self.set_cell_anchors(dtype, device)
anchors_over_all_feature_maps = self.cached_grid_anchors(grid_sizes, strides)
......
在之前提到，由于有 FPN 網(wǎng)絡(luò)，所以輸入 rpn 的是多個特征。為了方便介紹，以下都是以某一個特征進行描述，其他特征類似。
假設(shè)有 的特征，首先會計算這個特征相對于輸入圖像的下采樣倍數(shù) stride：
然后生成一個 大小的網(wǎng)格，每個格子長度為 stride，如下圖：
# AnchorGenerator.grid_anchors(...)
shifts_x = torch.arange(0, grid_width, dtype=torch.float32, device=device) * stride_width
shifts_y = torch.arange(0, grid_height, dtype=torch.float32, device=device) * stride_height
shift_y, shift_x = torch.meshgrid(shifts_y, shifts_x)
圖8
然后將 base_anchors 的中心從 移動到網(wǎng)格的點，且在網(wǎng)格的每個點都放置一組 base_anchors。這樣就在當(dāng)前 feature_map 上有了很多的 anchors。
需要特別說明，stride 代表網(wǎng)絡(luò)的感受野，網(wǎng)絡(luò)不可能檢測到比 feature_map 更密集的框了！所以才只會在網(wǎng)格中每個點設(shè)置 anchors（反過來說，如果在網(wǎng)格的兩個點之間設(shè)置 anchors，那么就對應(yīng) feature_map 中半個點，顯然不合理）。
# AnchorGenerator.grid_anchors(...)
anchors.append((shifts.view(-1, 1, 4) + base_anchors.view(1, -1, 4)).reshape(-1, 4))
圖9 （注：為了方便描述，這里只畫了3個anchor，實際每個點有9個anchor）
放置好 anchors 后，接下來就要調(diào)整網(wǎng)絡(luò)，使網(wǎng)絡(luò)輸出能夠判斷每個 anchor 是否有目標(biāo)，同時還要有 bounding box regression 需要的4個值 。
class RPNHead(nn.Module):
def __init__(self, in_channels, num_anchors):
super(RPNHead, self).__init__()
self.conv = nn.Conv2d(
in_channels, in_channels, kernel_size=3, stride=1, padding=1
)
self.cls_logits = nn.Conv2d(in_channels, num_anchors, kernel_size=1, stride=1)
self.bbox_pred = nn.Conv2d(
in_channels, num_anchors * 4, kernel_size=1, stride=1
)
def forward(self, x):
logits = []
bbox_reg = []
for feature in x:
t = F.relu(self.conv(feature))
logits.append(self.cls_logits(t))
bbox_reg.append(self.bbox_pred(t))
return logits, bbox_reg
假設(shè) feature 的大小 ，每個點 個 anchor。從 RPNHead 的代碼中可以明顯看到：
首先進行 3x3 卷積 然后對 feature 進行卷積，輸出 cls_logits 大小是 ，對應(yīng)每個 anchor 是否有目標(biāo)； 同時feature 進行卷積，輸出 bbox_pred 大小是 ，對應(yīng)每個點的4個框位置回歸信息 。
# RPNHead.__init__(...)
self.cls_logits = nn.Conv2d(in_channels, num_anchors, kernel_size=1, stride=1)
self.bbox_pred = nn.Conv2d(in_channels, num_anchors * 4, kernel_size=1, stride=1)
圖10（注：為了方便描述，這里只畫了3個anchor，實際每個點有9個anchor）
上述過程只是單個 feature_map 的處理流程。對于 FPN 網(wǎng)絡(luò)的輸出的多個大小不同的 feature_maps，每個特征圖都會按照上述過程計算 stride 和網(wǎng)格，并設(shè)置 anchors。當(dāng)處理完后獲得密密麻麻的各種 anchors 了。
接下來進入 RegionProposalNetwork 類：
# FasterRCNN.__init__(...)
rpn_pre_nms_top_n = dict(training=rpn_pre_nms_top_n_train, testing=rpn_pre_nms_top_n_test)
rpn_post_nms_top_n = dict(training=rpn_post_nms_top_n_train, testing=rpn_post_nms_top_n_test)
# rpn_anchor_generator 生成anchors
# rpn_head 調(diào)整feature_map獲得cls_logits+bbox_pred
rpn = RegionProposalNetwork(
rpn_anchor_generator, rpn_head,
rpn_fg_iou_thresh, rpn_bg_iou_thresh,
rpn_batch_size_per_image, rpn_positive_fraction,
rpn_pre_nms_top_n, rpn_post_nms_top_n, rpn_nms_thresh)RegionProposalNetwork 類的用是：
test 階段 ：計算有目標(biāo)的 anchor 并進行框回歸生成 proposals，然后 NMS train 階段 ：除了上面的作用，還計算 rpn loss
class RegionProposalNetwork(torch.nn.Module):
.......
def forward(self, images, features, targets=None):
features = list(features.values())
objectness, pred_bbox_deltas = self.head(features)
anchors = self.anchor_generator(images, features)
num_images = len(anchors)
num_anchors_per_level_shape_tensors = [o[0].shape for o in objectness]
num_anchors_per_level = [s[0] * s[1] * s[2] for s in num_anchors_per_level_shape_tensors]
objectness, pred_bbox_deltas = \
concat_box_prediction_layers(objectness, pred_bbox_deltas)
# apply pred_bbox_deltas to anchors to obtain the decoded proposals
# note that we detach the deltas because Faster R-CNN do not backprop through
# the proposals
proposals = self.box_coder.decode(pred_bbox_deltas.detach(), anchors)
proposals = proposals.view(num_images, -1, 4)
boxes, scores = self.filter_proposals(proposals, objectness, images.image_sizes, num_anchors_per_level)
losses = {}
if self.training:
assert targets is not None
labels, matched_gt_boxes = self.assign_targets_to_anchors(anchors, targets)
regression_targets = self.box_coder.encode(matched_gt_boxes, anchors)
loss_objectness, loss_rpn_box_reg = self.compute_loss(
objectness, pred_bbox_deltas, labels, regression_targets)
losses = {
"loss_objectness": loss_objectness,
"loss_rpn_box_reg": loss_rpn_box_reg,
}
return boxes, losses
具體來看，首先計算有目標(biāo)的 anchor 并進行框回歸生成 proposals ：
# RegionProposalNetwork.forward(...)
objectness, pred_bbox_deltas = self.head(features)
anchors = self.anchor_generator(images, features)
......
proposals = self.box_coder.decode(pred_bbox_deltas.detach(), anchors)
proposals = proposals.view(num_images, -1, 4)
然后依照 objectness 置信由大到小度排序（優(yōu)先提取更可能包含目標(biāo)的的），并 NMS，生成 boxes （即 NMS 后的 proposal boxes ） ：
# RegionProposalNetwork.forward(...)
boxes, scores = self.filter_proposals(proposals, objectness, images.image_sizes, num_anchors_per_level)
如果是訓(xùn)練階段，還要將 boxes 與 anchors 進行匹配，計算 cls_logits 的損失 loss_objectness，同時計算 bbox_pred 的損失 loss_rpn_box_reg。
在 RegionProposalNetwork 之后已經(jīng)生成了 boxes ，接下來就要提取 boxes 內(nèi)的特征進行 roi_pooling ：
roi_heads = RoIHeads(
# Box
box_roi_pool, box_head, box_predictor,
box_fg_iou_thresh, box_bg_iou_thresh,
box_batch_size_per_image, box_positive_fraction,
bbox_reg_weights,
box_score_thresh, box_nms_thresh, box_detections_per_img)
這里一點問題是如何計算 box 所屬的 feature_map：
對于原始 FasterRCNN，只在 backbone 的最后一層 feature_map 提取 box 對應(yīng)特征； 當(dāng)加入 FPN 后 backbone 會輸出多個特征圖，需要計算當(dāng)前 boxes 對應(yīng)于哪一個特征。
如下圖：
圖11
class MultiScaleRoIAlign(nn.Module):
......
def infer_scale(self, feature, original_size):
# type: (Tensor, List[int])
# assumption: the scale is of the form 2 ** (-k), with k integer
size = feature.shape[-2:]
possible_scales = torch.jit.annotate(List[float], [])
for s1, s2 in zip(size, original_size):
approx_scale = float(s1) / float(s2)
scale = 2 ** float(torch.tensor(approx_scale).log2().round())
possible_scales.append(scale)
assert possible_scales[0] == possible_scales[1]
return possible_scales[0]
def setup_scales(self, features, image_shapes):
# type: (List[Tensor], List[Tuple[int, int]])
assert len(image_shapes) != 0
max_x = 0
max_y = 0
for shape in image_shapes:
max_x = max(shape[0], max_x)
max_y = max(shape[1], max_y)
original_input_shape = (max_x, max_y)
scales = [self.infer_scale(feat, original_input_shape) for feat in features]
# get the levels in the feature map by leveraging the fact that the network always
# downsamples by a factor of 2 at each level.
lvl_min = -torch.log2(torch.tensor(scales[0], dtype=torch.float32)).item()
lvl_max = -torch.log2(torch.tensor(scales[-1], dtype=torch.float32)).item()
self.scales = scales
self.map_levels = initLevelMapper(int(lvl_min), int(lvl_max))
首先計算每個 feature_map 相對于網(wǎng)絡(luò)輸入 image 的下采樣倍率 scale。其中 infer_scale 函數(shù)采用如下的近似公式：
該公式相當(dāng)于做了一個簡單的映射，將不同的 feature_map 與 image 大小比映射到附近的尺度：
圖12
例如對于 FasterRCNN 實際值為：
之后設(shè)置 lvl_min=2 和 lvl_max=5：
# MultiScaleRoIAlign.setup_scales(...)
# get the levels in the feature map by leveraging the fact that the network always
# downsamples by a factor of 2 at each level.
lvl_min = -torch.log2(torch.tensor(scales[0], dtype=torch.float32)).item()
lvl_max = -torch.log2(torch.tensor(scales[-1], dtype=torch.float32)).item()
接著使用 FPN 原文中的公式計算 box 所在 anchor（其中 ， 為 box 面積）：
class LevelMapper(object)
def __init__(self, k_min, k_max, canonical_scale=224, canonical_level=4, eps=1e-6):
self.k_min = k_min # lvl_min=2
self.k_max = k_max # lvl_max=5
self.s0 = canonical_scale # 224
self.lvl0 = canonical_level # 4
self.eps = eps
def __call__(self, boxlists):
s = torch.sqrt(torch.cat([box_area(boxlist) for boxlist in boxlists]))
# Eqn.(1) in FPN paper
target_lvls = torch.floor(self.lvl0 + torch.log2(s / self.s0) + torch.tensor(self.eps, dtype=s.dtype))
target_lvls = torch.clamp(target_lvls, min=self.k_min, max=self.k_max)
return (target_lvls.to(torch.int64) - self.k_min).to(torch.int64)
其中 torch.clamp(input, min, max) → Tensor 函數(shù)的作用是截斷，防止越界：
可以看到，通過 LevelMapper 類將不同大小的 box 定位到某個 feature_map，如下圖。之后就是按照圖11中的流程進行 roi_pooling 操作。
圖13
在確定 proposal box 所屬 FPN 中哪個 feature_map 之后，接著來看 MultiScaleRoIAlign 如何進行 roi_pooling 操作：
class MultiScaleRoIAlign(nn.Module):
......
def forward(self, x, boxes, image_shapes):
# type: (Dict[str, Tensor], List[Tensor], List[Tuple[int, int]]) -> Tensor
x_filtered = []
for k, v in x.items():
if k in self.featmap_names:
x_filtered.append(v)
num_levels = len(x_filtered)
rois = self.convert_to_roi_format(boxes)
if self.scales is None:
self.setup_scales(x_filtered, image_shapes)
scales = self.scales
assert scales is not None
# 沒有 FPN 時，只有1/32的最后一個feature_map進行roi_pooling
if num_levels == 1:
return roi_align(
x_filtered[0], rois,
output_size=self.output_size,
spatial_scale=scales[0],
sampling_ratio=self.sampling_ratio
)
# 有 FPN 時，有4個feature_map進行roi_pooling
# 首先按照
mapper = self.map_levels
assert mapper is not None
levels = mapper(boxes)
num_rois = len(rois)
num_channels = x_filtered[0].shape[1]
dtype, device = x_filtered[0].dtype, x_filtered[0].device
result = torch.zeros(
(num_rois, num_channels,) + self.output_size,
dtype=dtype,
device=device,
)
tracing_results = []
for level, (per_level_feature, scale) in enumerate(zip(x_filtered, scales)):
idx_in_level = torch.nonzero(levels == level).squeeze(1)
rois_per_level = rois[idx_in_level]
result_idx_in_level = roi_align(
per_level_feature, rois_per_level,
output_size=self.output_size,
spatial_scale=scale, sampling_ratio=self.sampling_ratio)
if torchvision._is_tracing():
tracing_results.append(result_idx_in_level.to(dtype))
else:
result[idx_in_level] = result_idx_in_level
if torchvision._is_tracing():
result = _onnx_merge_levels(levels, tracing_results)
return result
在 MultiScaleRoIAlign.forward(...) 函數(shù)可以看到：
沒有 FPN 時，只有1/32的最后一個 feature_map 進行 roi_pooling
if num_levels == 1:
return roi_align(
x_filtered[0], rois,
output_size=self.output_size,
spatial_scale=scales[0],
sampling_ratio=self.sampling_ratio
)
有 FPN 時，有4個 的 feature maps 參加計算。首先計算每個每個 box 所屬哪個 feature map ，再在所屬 feature map 進行 roi_pooling
# 首先計算每個每個 box 所屬哪個 feature map
levels = mapper(boxes)
......
# 再在所屬 feature map 進行 roi_pooling
# 即 idx_in_level = torch.nonzero(levels == level).squeeze(1)
for level, (per_level_feature, scale) in enumerate(zip(x_filtered, scales)):
idx_in_level = torch.nonzero(levels == level).squeeze(1)
rois_per_level = rois[idx_in_level]
result_idx_in_level = roi_align(
per_level_feature, rois_per_level,
output_size=self.output_size,
spatial_scale=scale, sampling_ratio=self.sampling_ratio)
之后就獲得了所謂的 7x7 特征（在 FasterRCNN.__init__(...) 中設(shè)置了 output_size=7）。需要說明，原始 FasterRCNN 應(yīng)該是使用 roi_pooling，但是這里使用 roi_align 代替以提升檢測器性能。
對于 torchvision.ops.roi_align 函數(shù)輸入的參數(shù)，分別為：
per_level_feature 代表 FPN 輸出的某一 feature_map rois_per_level 為該特征 feature_map 對應(yīng)的所有 proposal boxes（之前計算 level得到） output_size=7 代表輸出為 7x7 spatial_scale 代表特征 feature_map 相對輸入 image 的下采樣尺度（如 1/4，1/8，...） sampling_ratio 為 roi_align 采樣率，有興趣的讀者請自行查閱 MaskRCNN 文章
接下來就是將特征轉(zhuǎn)為最后針對 box 的類別信息（如人、貓、狗、車）和進一步的框回歸信息。
class TwoMLPHead(nn.Module):
def __init__(self, in_channels, representation_size):
super(TwoMLPHead, self).__init__()
self.fc6 = nn.Linear(in_channels, representation_size)
self.fc7 = nn.Linear(representation_size, representation_size)
def forward(self, x):
x = x.flatten(start_dim=1)
x = F.relu(self.fc6(x))
x = F.relu(self.fc7(x))
return x
class FastRCNNPredictor(nn.Module):
def __init__(self, in_channels, num_classes):
super(FastRCNNPredictor, self).__init__()
self.cls_score = nn.Linear(in_channels, num_classes)
self.bbox_pred = nn.Linear(in_channels, num_classes * 4)
def forward(self, x):
if x.dim() == 4:
assert list(x.shape[2:]) == [1, 1]
x = x.flatten(start_dim=1)
scores = self.cls_score(x)
bbox_deltas = self.bbox_pred(x)
return scores, bbox_deltas
首先 TwoMLPHead 將 7x7 特征經(jīng)過兩個全連接層轉(zhuǎn)為 1024，然后 FastRCNNPredictor 將每個 box 對應(yīng)的 1024 維特征轉(zhuǎn)為 cls_score 和 bbox_pred :
圖14
顯然 cls_score 后接 softmax 即為類別概率，可以確定 box 的類別；在確定類別后，在 bbox_pred 中對應(yīng)類別的 4個值即為第二次 bounding box regression 需要的4個偏移值。
簡單的說，帶有FPN的FasterRCNN網(wǎng)絡(luò)結(jié)構(gòu)可以用下圖表示：
圖15
關(guān)于訓(xùn)練
FasterRCNN模型在兩處地方有損失函數(shù)：
在 RegionProposalNetwork 類，需要判別 anchor 中是否包含目標(biāo)從而生成 proposals，這里需要計算 loss 在 RoIHeads 類，對 roi_pooling 后的全連接生成的 cls_score 和 bbox_pred 進行訓(xùn)練，也需要計算 loss
首先來看 RegionProposalNetwork 類中的 assign_targets_to_anchors 函數(shù)。
def assign_targets_to_anchors(self, anchors, targets):
# type: (List[Tensor], List[Dict[str, Tensor]])
labels = []
matched_gt_boxes = []
for anchors_per_image, targets_per_image in zip(anchors, targets):
gt_boxes = targets_per_image["boxes"]
if gt_boxes.numel() == 0:
# Background image (negative example)
device = anchors_per_image.device
matched_gt_boxes_per_image = torch.zeros(anchors_per_image.shape, dtype=torch.float32, device=device)
labels_per_image = torch.zeros((anchors_per_image.shape[0],), dtype=torch.float32, device=device)
else:
match_quality_matrix = box_ops.box_iou(gt_boxes, anchors_per_image)
matched_idxs = self.proposal_matcher(match_quality_matrix)
# get the targets corresponding GT for each proposal
# NB: need to clamp the indices because we can have a single
# GT in the image, and matched_idxs can be -2, which goes
# out of bounds
matched_gt_boxes_per_image = gt_boxes[matched_idxs.clamp(min=0)]
labels_per_image = matched_idxs >= 0
labels_per_image = labels_per_image.to(dtype=torch.float32)
# Background (negative examples)
bg_indices = matched_idxs == self.proposal_matcher.BELOW_LOW_THRESHOLD
labels_per_image[bg_indices] = torch.tensor(0.0)
# discard indices that are between thresholds
inds_to_discard = matched_idxs == self.proposal_matcher.BETWEEN_THRESHOLDS
labels_per_image[inds_to_discard] = torch.tensor(-1.0)
labels.append(labels_per_image)
matched_gt_boxes.append(matched_gt_boxes_per_image)
return labels, matched_gt_boxes
當(dāng)圖像中沒有 gt_boxes 時，設(shè)置所有 anchor 都為 background（即 label 為 0）：
if gt_boxes.numel() == 0
# Background image (negative example)
device = anchors_per_image.device
matched_gt_boxes_per_image = torch.zeros(anchors_per_image.shape, dtype=torch.float32, device=device)
labels_per_image = torch.zeros((anchors_per_image.shape[0],), dtype=torch.float32, device=device)
當(dāng)圖像中有 gt_boxes 時，計算 anchor 與 gt_box 的 IOU：
選擇 IOU < 0.3 的 anchor 為 background，標(biāo)簽為 0
labels_per_image[bg_indices] = torch.tensor(0.0)
選擇 IOU > 0.7 的 anchor 為 foreground，標(biāo)簽為 1
labels_per_image = matched_idxs >= 0
忽略 0.3 < IOU < 0.7 的 anchor，不參與訓(xùn)練
從 FasterRCNN 類的 __init__ 函數(shù)默認參數(shù)就可以清晰的看到這一點：
rpn_fg_iou_thresh=0.7, rpn_bg_iou_thresh=0.3,
接著來看 RoIHeads 類中的 assign_targets_to_proposals 函數(shù)。
def assign_targets_to_proposals(self, proposals, gt_boxes, gt_labels):
# type: (List[Tensor], List[Tensor], List[Tensor])
matched_idxs = []
labels = []
for proposals_in_image, gt_boxes_in_image, gt_labels_in_image in zip(proposals, gt_boxes, gt_labels):
if gt_boxes_in_image.numel() == 0:
# Background image
device = proposals_in_image.device
clamped_matched_idxs_in_image = torch.zeros(
(proposals_in_image.shape[0],), dtype=torch.int64, device=device
)
labels_in_image = torch.zeros(
(proposals_in_image.shape[0],), dtype=torch.int64, device=device
)
else:
# set to self.box_similarity when https://github.com/pytorch/pytorch/issues/27495 lands
match_quality_matrix = box_ops.box_iou(gt_boxes_in_image, proposals_in_image)
matched_idxs_in_image = self.proposal_matcher(match_quality_matrix)
clamped_matched_idxs_in_image = matched_idxs_in_image.clamp(min=0)
labels_in_image = gt_labels_in_image[clamped_matched_idxs_in_image]
labels_in_image = labels_in_image.to(dtype=torch.int64)
# Label background (below the low threshold)
bg_inds = matched_idxs_in_image == self.proposal_matcher.BELOW_LOW_THRESHOLD
labels_in_image[bg_inds] = torch.tensor(0)
# Label ignore proposals (between low and high thresholds)
ignore_inds = matched_idxs_in_image == self.proposal_matcher.BETWEEN_THRESHOLDS
labels_in_image[ignore_inds] = torch.tensor(-1) # -1 is ignored by sampler
matched_idxs.append(clamped_matched_idxs_in_image)
labels.append(labels_in_image)
return matched_idxs, labels
與 assign_targets_to_anchors 不同，該函數(shù)設(shè)置：
box_fg_iou_thresh=0.5, box_bg_iou_thresh=0.5,
IOU > 0.5 的 proposal 為 foreground，標(biāo)簽為對應(yīng)的 class_id
labels_in_image = gt_labels_in_image[clamped_matched_idxs_in_image]
這里與上面不同：RegionProposalNetwork 只需要判斷 anchor 是否有目標(biāo)，正類別為1；RoIHeads 需要判斷 proposal 的具體類別，所以正類別為具體的 class_id。
IOU < 0.5 的為 background，標(biāo)簽為 0
labels_in_image[bg_inds] = torch.tensor(0)
寫在最后
本文簡要的介紹了 torchvision 中的 FasterRCNN 實現(xiàn)，并分析我認為重要的知識點。寫這篇文章的目的是為閱讀代碼困難的小伙伴做個指引，鼓勵入門新手能夠多看看代碼實現(xiàn)。若要真正的理解模型（不被面試官問住），要是要看代碼！
創(chuàng)作不易，不想被白嫖，求點贊、關(guān)注、收藏（作者）三連！
2020年11月國內(nèi)大數(shù)據(jù)競賽信息-獎池5000萬
刷爆網(wǎng)絡(luò)的動態(tài)條形圖，3行Python代碼就能搞定
↓掃描二維碼關(guān)注本號↓
求個在看！
評論
圖片
表情