def setup(self, bottom, top):
self._batch_size = 1
self._depth = 8
self._height = 240
self._width = 320
self.dataset = jhmdb('train', [self._height, self._width], split=1)
self.anchors, _, _ = self.dataset.get_anchors()
self.num_boxes = 64
def __init__(self):
self._batch_size = 1
self._depth = 8
self._height = 240
self._width = 320
self.dataset = jhmdb('trainval', [self._height, self._width], split=1)
self.anchors, self.valid_idx, self._anchor_dims = self.dataset.get_anchors(
)
def __init__(self, net, model):
self._batch_size = 1
self._depth = 8
self._height = 300
self._width = 400
self.dataset = jhmdb('val', [self._height, self._width], split=1)
caffe.set_mode_gpu()
self._net = caffe.Net(net, model, caffe.TEST)
def setup(self, bottom, top):
self._batch_size = 1
self._depth = 8
self._height = 300
self._width = 400
self.dataset = jhmdb('val', [self._height, self._width], split=1)
self.num_classes = self.dataset._num_classes - 1
self.anchors, self.valid_idx, self._anchor_dims = self.dataset.get_anchors(
)
def __init__(self, net, model):
self._batch_size = 1
self._depth = 8
self._height = 240
self._width = 320
self.dataset = jhmdb('val', [self._height, self._width], split=1)
self.anchors, self.valid_idx, self._anchor_dims = self.dataset.get_anchors()
caffe.set_mode_gpu()
self._net = caffe.Net(net, model, caffe.TEST)
def __init__(self, net, model):
self._batch_size = 1
self._depth = 8
self._height = 240
self._width = 320
self.dataset = jhmdb('val', [self._height, self._width],
split=1) #选取测试集，[height, width] = clip_shape
self.anchors, self.valid_idx, self._anchor_dims = self.dataset.get_anchors(
) #获取测试集anchors，valid_idx有效idx
caffe.set_mode_gpu()
self._net = caffe.Net(net, model, caffe.TEST) #网络前传
def __init__(self, net, model):
self._batch_size = 1
self._depth = 8
self._height = 300
self._width = 400
self._num_anchors = 10
self.dataset = jhmdb('train', [self._height, self._width], split=1)
self.base_anchors = np.load(
'/data/wjc/thumos14/cache/anchors_{}_{}.npy'.format(
self._depth, self._num_anchors)).transpose()
self._bottom_height = np.ceil(self._height / 16.0)
self._bottom_width = np.ceil(self._width / 16.0)
self._bottom_width = np.ceil(self._width / 16.0)
shift_x = np.arange(0, self._bottom_width)
shift_y = np.arange(0, self._bottom_height)
shift_x, shift_y = np.meshgrid(shift_x, shift_y)
shifts = np.vstack((shift_x.ravel(), shift_y.ravel(), shift_x.ravel(),
shift_y.ravel())).transpose()
# add A anchors (1, A, 4) to
# cell K shifts (K, 1, 4) to get
# shift anchors (K, A, 4)
# reshape to (K*A, 4) shifted anchors
A = self._num_anchors
K = shifts.shape[0]
all_anchors = (self.base_anchors.reshape((1, A, 4)) + shifts.reshape(
(1, K, 4)).transpose((1, 0, 2)))
all_anchors = all_anchors.reshape((K * A, 4))
total_anchors = int(K * A)
# only keep anchors inside the image
inds_inside = np.where(
(all_anchors[:, 0] >= 0) & (all_anchors[:, 1] >= 0)
& (all_anchors[:, 2] < self._bottom_width) & # width
(all_anchors[:, 3] < self._bottom_width) # height
)[0]
self.anchors = all_anchors[inds_inside, :]
caffe.set_mode_gpu()
self._net = caffe.Net(net, model, caffe.TEST)
self._list = []
from dataset.jhmdb import jhmdb
j = jhmdb(name='train', clip_shape=[240, 320], split=1)
for i in xrange(4, 17):
j.cluster_bboxes(anchors=i)
