Python torch.nn.functional.affine_grid() Examples

The following are 30 code examples of torch.nn.functional.affine_grid(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module torch.nn.functional , or try the search function .
Example #1
Source File: net_utils.py    From faster-rcnn.pytorch with MIT License 7 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #2
Source File: DDPAE_utils.py    From DDPAE-video-prediction with MIT License 7 votes vote down vote up 
def image_to_object(images, pose, object_size):
  '''
  Inverse pose, crop and transform image patches.
  param images: (... x C x H x W) tensor
  param pose: (N x 3) tensor
  '''
  N, pose_size = pose.size()
  n_channels, H, W = images.size()[-3:]
  images = images.view(N, n_channels, H, W)
  if pose_size == 3:
    transformer_inv = expand_pose(pose_inv(pose))
  elif pose_size == 6:
    transformer_inv = pose_inv_full(pose)

  grid = F.affine_grid(transformer_inv,
                       torch.Size((N, n_channels, object_size, object_size)))
  obj = F.grid_sample(images, grid)
  return obj
Example #3
Source File: DDPAE_utils.py    From DDPAE-video-prediction with MIT License 7 votes vote down vote up 
def object_to_image(objects, pose, image_size):
  '''
  param images: (N x C x H x W) tensor
  param pose: (N x 3) tensor
  '''
  N, pose_size = pose.size()
  _, n_channels, _, _ = objects.size()
  if pose_size == 3:
    transformer = expand_pose(pose)
  elif pose_size == 6:
    transformer = pose.view(N, 2, 3)

  grid = F.affine_grid(transformer,
                       torch.Size((N, n_channels, image_size, image_size)))
  components = F.grid_sample(objects, grid)
  return components
Example #4
Source File: net.py    From PANet with MIT License 6 votes vote down vote up 
def affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #5
Source File: point_sample.py    From mmdetection with Apache License 2.0 6 votes vote down vote up 
def generate_grid(num_grid, size, device):
    """Generate regular square grid of points in [0, 1] x [0, 1] coordinate
    space.

    Args:
        num_grid (int): The number of grids to sample, one for each region.
        size (tuple(int, int)): The side size of the regular grid.
        device (torch.device): Desired device of returned tensor.

    Returns:
        (torch.Tensor): A tensor of shape (num_grid, size[0]*size[1], 2) that
            contains coordinates for the regular grids.
    """

    affine_trans = torch.tensor([[[1., 0., 0.], [0., 1., 0.]]], device=device)
    grid = F.affine_grid(
        affine_trans, torch.Size((1, 1, *size)), align_corners=False)
    grid = normalize(grid)
    return grid.view(1, -1, 2).expand(num_grid, -1, -1)
Example #6
Source File: warp.py    From open-vot with MIT License 6 votes vote down vote up 
def crop_tensor(image, center, size, padding='avg', out_size=None):
    assert out_size is not None
    img_sz = torch.tensor(image.size()[:-3:-1]).to(image.device).float()

    # calculate padding
    corners = torch.cat((center - size / 2, center + size / 2))
    pads = torch.cat((-corners[:2], corners[2:] - img_sz))
    npad = max(0, pads.max().item())

    if npad > 0 and padding == 'avg':
        avg_chan = image.view(3, -1).mean(dim=1).view(1, 3, 1, 1)
        image -= avg_chan

    out_size = torch.Size((1, 1, out_size, out_size))

    theta = encode_theta(center, size, 0, img_sz)
    grid = F.affine_grid(theta, out_size)
    patch = F.grid_sample(image, grid)

    if npad > 0 and padding == 'avg':
        patch += avg_chan

    return patch
Example #7
Source File: net_utils.py    From Context-aware-ZSR with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #8
Source File: net.py    From Context-aware-ZSR with MIT License 6 votes vote down vote up 
def affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #9
Source File: net_utils.py    From Detectron.pytorch with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #10
Source File: net_utils.py    From pytorch-detect-to-track with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #11
Source File: net_utils.py    From PANet with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #12
Source File: point_sample.py    From mmcv with Apache License 2.0 6 votes vote down vote up 
def generate_grid(num_grid, size, device):
    """Generate regular square grid of points in [0, 1] x [0, 1] coordinate
    space.

    Args:
        num_grid (int): The number of grids to sample, one for each region.
        size (tuple(int, int)): The side size of the regular grid.
        device (torch.device): Desired device of returned tensor.

    Returns:
        (torch.Tensor): A tensor of shape (num_grid, size[0]*size[1], 2) that
            contains coordinates for the regular grids.
    """

    affine_trans = torch.tensor([[[1., 0., 0.], [0., 1., 0.]]], device=device)
    grid = F.affine_grid(
        affine_trans, torch.Size((1, 1, *size)), align_corners=False)
    grid = normalize(grid)
    return grid.view(1, -1, 2).expand(num_grid, -1, -1)
Example #13
Source File: dense_pdd_net_v01.py    From pdd_net with Apache License 2.0 6 votes vote down vote up 
def augmentAffine(img_in, seg_in, strength=0.05):
    """
    3D affine augmentation on image and segmentation mini-batch on GPU.
    (affine transf. is centered: trilinear interpolation and zero-padding used for sampling)
    :input: img_in batch (torch.cuda.FloatTensor), seg_in batch (torch.cuda.LongTensor)
    :return: augmented BxCxTxHxW image batch (torch.cuda.FloatTensor), augmented BxTxHxW seg batch (torch.cuda.LongTensor)
    """
    B,C,D,H,W = img_in.size()
    affine_matrix = (torch.eye(3,4).unsqueeze(0) + torch.randn(B, 3, 4) * strength).to(img_in.device)

    meshgrid = F.affine_grid(affine_matrix,torch.Size((B,1,D,H,W)))

    img_out = F.grid_sample(img_in, meshgrid,padding_mode='border')
    seg_out = F.grid_sample(seg_in.float().unsqueeze(1), meshgrid, mode='nearest').long().squeeze(1)

    return img_out, seg_out
Example #14
Source File: net_utils.py    From pcl.pytorch with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #15
Source File: net_utils.py    From Distilling-Object-Detectors with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #16
Source File: net.py    From FPN-Pytorch with MIT License 6 votes vote down vote up 
def affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #17
Source File: net_utils.py    From bottom-up-features with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #18
Source File: net_utils.py    From FPN-Pytorch with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #19
Source File: net.py    From pcl.pytorch with MIT License 6 votes vote down vote up 
def affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #20
Source File: net_utils.py    From FPN_Pytorch with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #21
Source File: net_utils.py    From One-Shot-Object-Detection with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #22
Source File: two_algo_face_rotator.py    From talking-head-anime-demo with MIT License 6 votes vote down vote up 
def forward(self, image: Tensor, pose: Tensor):
        n = image.size(0)
        c = image.size(1)
        h = image.size(2)
        w = image.size(3)

        pose = pose.unsqueeze(2).unsqueeze(3)
        pose = pose.expand(pose.size(0), pose.size(1), image.size(2), image.size(3))
        x = torch.cat([image, pose], dim=1)
        y = self.main_body(x)

        color_change = self.pumarola_color_change(y)
        alpha_mask = self.pumarola_alpha_mask(y)
        color_changed = alpha_mask * image + (1 - alpha_mask) * color_change

        grid_change = torch.transpose(self.zhou_grid_change(y).view(n, 2, h * w), 1, 2).view(n, h, w, 2)
        device = self.zhou_grid_change.weight.device
        identity = torch.Tensor([[1, 0, 0], [0, 1, 0]]).to(device).unsqueeze(0).repeat(n, 1, 1)
        base_grid = affine_grid(identity, [n, c, h, w], align_corners=self.align_corners)
        grid = base_grid + grid_change
        resampled = grid_sample(image, grid, mode='bilinear', padding_mode='border', align_corners=self.align_corners)

        return [color_changed, resampled, color_change, alpha_mask, grid_change, grid]
Example #23
Source File: stn.py    From LaSO with BSD 3-Clause "New" or "Revised" License 6 votes vote down vote up 
def forward(self, x):
        #
        # Calculate the transform
        #
        xs = self.localization(x)
        xs = xs.view(-1, 32*7*7)

        theta = self.fc_loc(xs)
        theta = theta.view(-1, 2, 3)

        grid = F.affine_grid(theta, x.size())

        #
        # transform the input
        #
        x = F.grid_sample(x, grid)

        return x
Example #24
Source File: basic_utils_batch.py    From landmark-detection with MIT License 6 votes vote down vote up 
def warp_feature_batch(feature, pts_location, patch_size):
  # feature must be [1,C,H,W] and pts_location must be [Num-Pts, (x,y)]
  _, C, H, W = list(feature.size())
  num_pts = pts_location.size(0)
  assert isinstance(patch_size, int) and feature.size(0) == 1 and pts_location.size(1) == 2, 'The shapes of feature or points are not right : {} vs {}'.format(feature.size(), pts_location.size())
  assert W > 1 and H > 1, 'To guarantee normalization {}, {}'.format(W, H)

  def normalize(x, L):
    return -1. + 2. * x / (L-1)

  crop_box = torch.cat([pts_location-patch_size, pts_location+patch_size], 1)
  crop_box[:, [0,2]] = normalize(crop_box[:, [0,2]], W)
  crop_box[:, [1,3]] = normalize(crop_box[:, [1,3]], H)
 
  affine_parameter = [(crop_box[:,2]-crop_box[:,0])/2, crop_box[:,0]*0, (crop_box[:,2]+crop_box[:,0])/2,
                      crop_box[:,0]*0, (crop_box[:,3]-crop_box[:,1])/2, (crop_box[:,3]+crop_box[:,1])/2]
  #affine_parameter = [(crop_box[:,2]-crop_box[:,0])/2, MU.np2variable(torch.zeros(num_pts),feature.is_cuda,False), (crop_box[:,2]+crop_box[:,0])/2,
  #                    MU.np2variable(torch.zeros(num_pts),feature.is_cuda,False), (crop_box[:,3]-crop_box[:,1])/2, (crop_box[:,3]+crop_box[:,1])/2]
  theta = torch.stack(affine_parameter, 1).view(num_pts, 2, 3)
  feature = feature.expand(num_pts,C, H, W)
  grid_size = torch.Size([num_pts, 1, 2*patch_size+1, 2*patch_size+1])
  grid = F.affine_grid(theta, grid_size)
  sub_feature = F.grid_sample(feature, grid)
  return sub_feature
Example #25
Source File: net_utils.py    From RFCN_CoupleNet.pytorch with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #26
Source File: net_utils.py    From CIOD with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):
    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([ \
        (x2 - x1) / (width - 1),
        zero,
        (x1 + x2 - width + 1) / (width - 1),
        zero,
        (y2 - y1) / (height - 1),
        (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #27
Source File: net_utils.py    From fpn.pytorch with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid
Example #28
Source File: model_utils.py    From landmark-detection with MIT License 6 votes vote down vote up 
def roi_pooling(input, rois, size=(7,7)):
  assert rois.dim() == 2 and rois.size(1) == 5, 'rois shape is wrong : {}'.format(rois.size())
  output = []
  num_rois = rois.size(0)
  size = np.array(size)
  spatial_size = np.array([input.size(3), input.size(2)])
  for i in range(num_rois):
    roi = variable2np(rois[i])
    im_idx = int(roi[0])
    theta = utils.crop2affine(spatial_size, roi[1:])
    theta = np2variable(theta, input.is_cuda).unsqueeze(0)
    grid_size = torch.Size([1, 3, int(size[1]), int(size[0])])
    grid = F.affine_grid(theta, grid_size)
    roi_feature = F.grid_sample(input.narrow(0, im_idx, 1), grid)
    output.append( roi_feature )
  return torch.cat(output, 0)
Example #29
Source File: gen_mean_face.py    From landmark-detection with MIT License 6 votes vote down vote up 
def face_align(face, point, target):
  spatial_size = np.array(face.size)
  point, target = point.copy(), target.copy()
  point[:,0] = normalize(spatial_size[0], point[:,0])
  point[:,1] = normalize(spatial_size[1], point[:,1])
  target[:,0] = normalize(spatial_size[0], target[:,0])
  target[:,1] = normalize(spatial_size[1], target[:,1])
  x, residual, rank, s = np.linalg.lstsq(target, point)
  theta = x.T[:2,:]
  theta = np2variable(theta).unsqueeze(0)
  image = np.array(face.copy()).transpose(2, 0, 1)
  image_var = np2variable(image, False).unsqueeze(0)
  grid_size = torch.Size([1, 3, int(spatial_size[1]), int(spatial_size[0])])
  grid = F.affine_grid(theta, grid_size)
  aligned_image = F.grid_sample(image_var, grid)
  aligned_image = aligned_image.data.numpy().squeeze()
  aligned_image = aligned_image.transpose(1, 2, 0)
  aligned_image = Image.fromarray(np.uint8(aligned_image))
  return aligned_image
Example #30
Source File: utils.py    From NeuralBabyTalk with MIT License 6 votes vote down vote up 
def _affine_grid_gen(rois, input_size, grid_size):

    rois = rois.detach()
    x1 = rois[:, 1::4] / 16.0
    y1 = rois[:, 2::4] / 16.0
    x2 = rois[:, 3::4] / 16.0
    y2 = rois[:, 4::4] / 16.0

    height = input_size[0]
    width = input_size[1]

    zero = Variable(rois.data.new(rois.size(0), 1).zero_())
    theta = torch.cat([\
      (x2 - x1) / (width - 1),
      zero,
      (x1 + x2 - width + 1) / (width - 1),
      zero,
      (y2 - y1) / (height - 1),
      (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)

    grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))

    return grid