roidb是比较复杂的数据结构，存放了数据集的roi信息。原始的roidb来自数据集，在trian.py的get_training_roidb(imdb)函数进行了水平翻转扩充数量，然后prepare_roidb(imdb)【定义在roidb.py】为roidb添加了一些说明性的属性。
在这里暂时记录下roidb的结构信息，后面继续看的时候可能会有些修正：
roidb是由字典组成的list，roidb[img_index]包含了该图片索引所包含到roi信息，下面以roidb[img_index]为例说明：
| roidb[img_index]包含的key, | value |
| boxes | box位置信息，box_num*4的np array |
| gt_overlaps | 所有box在不同类别的得分，box_num*class_num矩阵 |
| gt_classes | 所有box的真实类别，box_num长度的list |
| flipped | 是否翻转 |
| image | 该图片的路径，字符串 |
| width | 图片的宽 |
| height | 图片的高 |
| max_overlaps | 每个box的在所有类别的得分最大值，box_num长度 |
| max_classes | 每个box的得分最高所对应的类，box_num长度 |
| bbox_targets | 每个box的类别，以及与最接近的gt-box的4个方位偏移 （共5列） |
def add_bbox_regression_targets(roidb): """Add information needed to train bounding-box regressors.""" assert len(roidb) > 0 assert 'max_classes' in roidb[0], 'Did you call prepare_roidb first?' num_images = len(roidb) # Infer number of classes from the number of columns in gt_overlaps # 类别数，roidb[0]对应第0号图片上的roi,shape[1]多少列表示roi属于不同类上的概率 num_classes = roidb[0]['gt_overlaps'].shape[1] for im_i in xrange(num_images): rois = roidb[im_i]['boxes'] max_overlaps = roidb[im_i]['max_overlaps'] max_classes = roidb[im_i]['max_classes'] # bbox_targets：每个box的类别，以及与最接近的gt-box的4个方位偏移 roidb[im_i]['bbox_targets'] = \ _compute_targets(rois, max_overlaps, max_classes) # 这里config是false if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED: # Use fixed / precomputed "means" and "stds" instead of empirical values # 使用固定的均值和方差代替经验值 means = np.tile( np.array(cfg.TRAIN.BBOX_NORMALIZE_MEANS), (num_classes, 1)) stds = np.tile( np.array(cfg.TRAIN.BBOX_NORMALIZE_STDS), (num_classes, 1)) else: # Compute values needed for means and stds # 计算所需的均值和方差 # var(x) = E(x^2) - E(x)^2 # 计数各个类别出现box的数量 class_counts = np.zeros((num_classes, 1)) + cfg.EPS #加上cfg.EPS防止除0出错 # 21类*4个位置，如果出现box的类别与其中某一类相同，将该box的4个target加入4个列元素中 sums = np.zeros((num_classes, 4)) # 21类*4个位置，如果出现box的类别与其中某一类相同，将该box的4个target的平方加入4个列元素中 squared_sums = np.zeros((num_classes, 4)) for im_i in xrange(num_images): targets = roidb[im_i]['bbox_targets'] for cls in xrange(1, num_classes): cls_inds = np.where(targets[:, 0] == cls)[0] # box的类别与该类匹配，计入 if cls_inds.size > 0: class_counts[cls] += cls_inds.size sums[cls, :] += targets[cls_inds, 1:].sum(axis=0) squared_sums[cls, :] += \ (targets[cls_inds, 1:] ** 2).sum(axis=0) means = sums / class_counts # 均值 stds = np.sqrt(squared_sums / class_counts - means ** 2) #标准差 print 'bbox target means:' print means print means[1:, :].mean(axis=0) # ignore bg class print 'bbox target stdevs:' print stds print stds[1:, :].mean(axis=0) # ignore bg class # Normalize targets # 对每一box归一化target if cfg.TRAIN.BBOX_NORMALIZE_TARGETS: print "Normalizing targets" for im_i in xrange(num_images): targets = roidb[im_i]['bbox_targets'] for cls in xrange(1, num_classes): cls_inds = np.where(targets[:, 0] == cls)[0] roidb[im_i]['bbox_targets'][cls_inds, 1:] -= means[cls, :] roidb[im_i]['bbox_targets'][cls_inds, 1:] /= stds[cls, :] else: print "NOT normalizing targets" # 均值和方差也用于预测 # These values will be needed for making predictions # (the predicts will need to be unnormalized and uncentered) return means.ravel(), stds.ravel() # ravel()排序拉成一维 def _compute_targets(rois, overlaps, labels): # 参数rois只含有当前图片的box信息 """Compute bounding-box regression targets for an image.""" # Indices目录 of ground-truth ROIs # ground-truth ROIs gt_inds = np.where(overlaps == 1)[0] if len(gt_inds) == 0: # Bail if the image has no ground-truth ROIs # 不存在gt ROI，返回空数组 return np.zeros((rois.shape[0], 5), dtype=np.float32) # Indices of examples for which we try to make predictions # BBOX阈值，只有ROI与gt的重叠度大于阈值，这样的ROI才能用作bb回归的训练样本 ex_inds = np.where(overlaps >= cfg.TRAIN.BBOX_THRESH)[0] # Get IoU overlap between each ex ROI and gt ROI # 计算ex ROI and gt ROI的IoU ex_gt_overlaps = bbox_overlaps( # 变数据格式为float np.ascontiguousarray(rois[ex_inds, :], dtype=np.float), np.ascontiguousarray(rois[gt_inds, :], dtype=np.float)) # Find which gt ROI each ex ROI has max overlap with: # this will be the ex ROI's gt target # 这里每一行代表一个ex_roi,列代表gt_roi,元素数值代表两者的IoU gt_assignment = ex_gt_overlaps.argmax(axis=1) #按行求最大，返回索引. gt_rois = rois[gt_inds[gt_assignment], :] #每个ex_roi对应的gt_rois，与下面ex_roi数量相同 ex_rois = rois[ex_inds, :] targets = np.zeros((rois.shape[0], 5), dtype=np.float32) targets[ex_inds, 0] = labels[ex_inds] #第一个元素是label targets[ex_inds, 1:] = bbox_transform(ex_rois, gt_rois) #后4个元素是ex_box与gt_box的4个方位的偏移 return targets
下面解读一下这两个函数。
1. _compute_targets(rois, overlaps, labels)
这个函数用来计算roi的偏移量。基本的步骤就是先确认是否含有ground-truth rois,主要通过 ground-truth ROIs的overlaps=1来确认。
然后找到重叠度大于一定阈值的box，再进行计算。
对于满足条件的box,会调用程序bbox_overlaps重新计算box对应于ground-truth box的重叠度，根据最大的重叠度找对应的ground truth box.
这样就可以利用 fast_rcnn.bbox_transform 的bbox_transform计算4个偏移（分别是中心点的x,y坐标，w,d长度偏移）。
输出的是一个二维数组，横坐标是盒子的序号，纵坐标是5维，第一维是类别，第二维到第五维为偏移。
bbox_overlaps的代码如下：
def bbox_overlaps(
np.ndarray[DTYPE_t, ndim=2] boxes,
np.ndarray[DTYPE_t, ndim=2] query_boxes):
"""
Parameters
----------
boxes: (N, 4) ndarray of float
query_boxes: (K, 4) ndarray of float
Returns
-------
overlaps: (N, K) ndarray of overlap between boxes and query_boxes
"""
cdef unsigned int N = boxes.shape[0]
cdef unsigned int K = query_boxes.shape[0]
cdef np.ndarray[DTYPE_t, ndim=2] overlaps = np.zeros((N, K), dtype=DTYPE)
cdef DTYPE_t iw, ih, box_area
cdef DTYPE_t ua
cdef unsigned int k, n
for k in range(K):
box_area = (
(query_boxes[k, 2] - query_boxes[k, 0] + 1) *
(query_boxes[k, 3] - query_boxes[k, 1] + 1)
)
for n in range(N):
iw = (
min(boxes[n, 2], query_boxes[k, 2]) -
max(boxes[n, 0], query_boxes[k, 0]) + 1
)
if iw > 0:
ih = (
min(boxes[n, 3], query_boxes[k, 3]) -
max(boxes[n, 1], query_boxes[k, 1]) + 1
)
if ih > 0:
ua = float(
(boxes[n, 2] - boxes[n, 0] + 1) *
(boxes[n, 3] - boxes[n, 1] + 1) +
box_area - iw * ih
)
overlaps[n, k] = iw * ih / ua
return overlaps2. add_bbox_regression_targets
主要两个两件事： 1. 确定roidb每个图片的box的回归偏移量bbox_targets
2. 对于所有的类别，计算偏移量的均值和方差，这样输出的矩阵是二维，行坐标是种类（这里是21类），纵坐标是偏移量（这里是4）.
并且在需要正则化目标项（即cfg.TRAIN.BBOX_NORMALIZE_TARGETS=true）时，使每个偏移都减去均值除以标准差。并返回均值和方差的折叠一维向量，
用于预测（即逆操作，去正则化，则中心化）。
参考：
1476
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
