Python nd.abs方法代碼示例

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def _reg_loss(regr, gt_regr, mask):
  ''' L1 regression loss
    Arguments:
      regr (batch x max_objects x dim)
      gt_regr (batch x max_objects x dim)
      mask (batch x max_objects)
  '''
  num = mask.astype('float32').sum()
  mask = mask.expand_dims(2).broadcast_like(gt_regr).astype('float32')

  regr = regr * mask
  gt_regr = gt_regr * mask

  t = nd.abs(regr - gt_regr)
  regt_loss = smooth_l1(t).sum()
  regr_loss = regr_loss / (num + 1e-4)
  return regr_loss 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def Route(self, x):
        # b_mat = nd.repeat(self.b_mat.data(), repeats=x.shape[0], axis=0)#nd.stop_gradient(nd.repeat(self.b_mat.data(), repeats=x.shape[0], axis=0))
        b_mat = nd.zeros((x.shape[0],1,self.num_cap, self.num_locations), ctx=x.context)
        x_expand = nd.expand_dims(nd.expand_dims(x, axis=2),2)
        w_expand = nd.repeat(nd.expand_dims(self.w_ij.data(x.context),axis=0), repeats=x.shape[0], axis=0)
        u_ = w_expand*x_expand
        # u_ = nd.abs(w_expand - x_expand)
        u = nd.sum(u_, axis = 1)
        u_no_gradient = nd.stop_gradient(u)
        for i in range(self.route_num):
            c_mat = nd.softmax(b_mat, axis=2)
            if i == self.route_num -1:
                s = nd.sum(u * c_mat, axis=-1)
            else:
                s = nd.sum(u_no_gradient * c_mat, axis=-1)
            v = squash(s, 1)
            v1 = nd.expand_dims(v, axis=-1)
            if i != self.route_num - 1:
                update_term = nd.sum(u_no_gradient*v1, axis=1, keepdims=True)
                b_mat = b_mat + update_term
        return v 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def hybrid_forward(self, F, pred, label, mask, sample_weight=None):
        label = _reshape_like(F, label, pred)
        loss = F.abs(label * mask - pred * mask)
        loss = _apply_weighting(F, loss, self._weight, sample_weight)
        norm = F.sum(mask).clip(1, 1e30)
        return F.sum(loss) / norm 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def weight_l1_loss(self, F, pred_loc, label_loc, loss_weight):
        """Compute weight_l1_loss"""
        pred_loc = pred_loc.reshape((self.b, 4, -1, self.h, self.w))
        diff = F.abs((pred_loc - label_loc))
        diff = F.sum(diff, axis=1).reshape((self.b, -1, self.h, self.w))
        loss = diff * loss_weight
        return F.sum(loss)/self.b 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def extract_multi_position_matrix_nd(bbox):
    bbox = nd.transpose(bbox, axes=(1, 0, 2))
    xmin, ymin, xmax, ymax = nd.split(data=bbox, num_outputs=4, axis=2)
    # [num_fg_classes, num_boxes, 1]
    bbox_width = xmax - xmin + 1.
    bbox_height = ymax - ymin + 1.
    center_x = 0.5 * (xmin + xmax)
    center_y = 0.5 * (ymin + ymax)
    # [num_fg_classes, num_boxes, num_boxes]
    delta_x = nd.broadcast_minus(lhs=center_x, 
        rhs=nd.transpose(center_x, axes=(0, 2, 1)))
    delta_x = nd.broadcast_div(delta_x, bbox_width)
    delta_x = nd.log(nd.maximum(nd.abs(delta_x), 1e-3))

    delta_y = nd.broadcast_minus(lhs=center_y,
        rhs=nd.transpose(center_y, axes=(0, 2, 1)))
    delta_y = nd.broadcast_div(delta_y, bbox_height)
    delta_y = nd.log(nd.maximum(nd.abs(delta_y), 1e-3))

    delta_width = nd.broadcast_div(lhs=bbox_width, 
        rhs=nd.transpose(bbox_width, axes=(0, 2, 1)))
    delta_width = nd.log(delta_width)

    delta_height = nd.broadcast_div(lhs=bbox_height,
        rhs=nd.transpose(bbox_height, axes=(0, 2, 1)))
    delta_height = nd.log(delta_height)
    concat_list = [delta_x, delta_y, delta_width, delta_height]
    for idx, sym in enumerate(concat_list):
        concat_list[idx] = nd.expand_dims(sym, axis=3)
    position_matrix = nd.concat(*concat_list, dim=3)
    return position_matrix 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def _slow_reg_loss(regr, gt_regr, mask):
    num  = mask.float().sum()
    mask = mask.unsqueeze(2).expand_as(gt_regr)

    regr    = regr[mask]
    gt_regr = gt_regr[mask]

    t = nd.abs(regr - gt_regr)
    regt_loss = smooth_l1(t).sum()
    regr_loss = regr_loss / (num + 1e-4)
    return regr_loss 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def forward(self, output, mask, ind, target):
    pred = _tranpose_and_gather_feat(output, ind)
    pred = pred.swapaxes(dim1 = 0, dim2 = 1)
    mask = mask.expand_dims(axis = 2).broadcast_like(pred).astype('float32')
    loss = nd.abs(pred*mask - target*mask).sum()
    loss = loss / (mask.sum() + 1e-4)
    return loss 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def compute_res_loss(output, target, maski):
    diff = nd.abs(output * maski - target * maski)
    return nd.smooth_l1(diff).mean() 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def forward(self, cls_pred, box_pred, cls_target, box_target):
        """Compute loss in entire batch across devices."""
        # require results across different devices at this time
        cls_pred, box_pred, cls_target, box_target = [_as_list(x) \
            for x in (cls_pred, box_pred, cls_target, box_target)]
        # cross device reduction to obtain positive samples in entire batch
        num_pos = []
        for cp, bp, ct, bt in zip(*[cls_pred, box_pred, cls_target, box_target]):
            pos_samples = (ct > 0)
            num_pos.append(pos_samples.sum())
        num_pos_all = sum([p.asscalar() for p in num_pos])
        if num_pos_all < 1:
            # no positive samples found, return dummy losses
            return nd.zeros((1,)), nd.zeros((1,)), nd.zeros((1,))

        # compute element-wise cross entropy loss and sort, then perform negative mining
        cls_losses = []
        box_losses = []
        sum_losses = []
        for cp, bp, ct, bt in zip(*[cls_pred, box_pred, cls_target, box_target]):
            pred = nd.log_softmax(cp, axis=-1)
            pos = ct > 0
            cls_loss = -nd.pick(pred, ct, axis=-1, keepdims=False)
            rank = (cls_loss * (pos - 1)).argsort(axis=1).argsort(axis=1)
            hard_negative = rank < (pos.sum(axis=1) * self._negative_mining_ratio).expand_dims(-1)
            # mask out if not positive or negative
            cls_loss = nd.where((pos + hard_negative) > 0, cls_loss, nd.zeros_like(cls_loss))
            cls_losses.append(nd.sum(cls_loss, axis=0, exclude=True) / num_pos_all)

            bp = _reshape_like(nd, bp, bt)
            box_loss = nd.abs(bp - bt)
            box_loss = nd.where(box_loss > self._rho, box_loss - 0.5 * self._rho,
                                (0.5 / self._rho) * nd.square(box_loss))
            # box loss only apply to positive samples
            box_loss = box_loss * pos.expand_dims(axis=-1)
            box_losses.append(nd.sum(box_loss, axis=0, exclude=True) / num_pos_all)
            sum_losses.append(cls_losses[-1] + self._lambd * box_losses[-1])

        return sum_losses, cls_losses, box_losses 

# 需要導入模塊: from mxnet import nd [as 別名]
# 或者: from mxnet.nd import abs [as 別名]
def forward(self, cls_pred, box_pred, cls_target, box_target):
        """Compute loss in entire batch across devices."""
        # require results across different devices at this time
        cls_pred, box_pred, cls_target, box_target = [_as_list(x) \
            for x in (cls_pred, box_pred, cls_target, box_target)]
        # cross device reduction to obtain positive samples in entire batch
        num_pos = []
        for cp, bp, ct, bt in zip(*[cls_pred, box_pred, cls_target, box_target]):
            pos_samples = (ct > 0)
            num_pos.append(pos_samples.sum())
        num_pos_all = sum([p.asscalar() for p in num_pos])
        if num_pos_all < 1 and self._min_hard_negatives < 1:
            # no positive samples and no hard negatives, return dummy losses
            cls_losses = [nd.sum(cp * 0) for cp in cls_pred]
            box_losses = [nd.sum(bp * 0) for bp in box_pred]
            sum_losses = [nd.sum(cp * 0) + nd.sum(bp * 0) for cp, bp in zip(cls_pred, box_pred)]
            return sum_losses, cls_losses, box_losses


        # compute element-wise cross entropy loss and sort, then perform negative mining
        cls_losses = []
        box_losses = []
        sum_losses = []
        for cp, bp, ct, bt in zip(*[cls_pred, box_pred, cls_target, box_target]):
            pred = nd.log_softmax(cp, axis=-1)
            pos = ct > 0
            cls_loss = -nd.pick(pred, ct, axis=-1, keepdims=False)
            rank = (cls_loss * (pos - 1)).argsort(axis=1).argsort(axis=1)
            hard_negative = rank < nd.maximum(self._min_hard_negatives, pos.sum(axis=1)
                                              * self._negative_mining_ratio).expand_dims(-1)
            # mask out if not positive or negative
            cls_loss = nd.where((pos + hard_negative) > 0, cls_loss, nd.zeros_like(cls_loss))
            cls_losses.append(nd.sum(cls_loss, axis=0, exclude=True) / max(1., num_pos_all))

            bp = _reshape_like(nd, bp, bt)
            box_loss = nd.abs(bp - bt)
            box_loss = nd.where(box_loss > self._rho, box_loss - 0.5 * self._rho,
                                (0.5 / self._rho) * nd.square(box_loss))
            # box loss only apply to positive samples
            box_loss = box_loss * pos.expand_dims(axis=-1)
            box_losses.append(nd.sum(box_loss, axis=0, exclude=True) / max(1., num_pos_all))
            sum_losses.append(cls_losses[-1] + self._lambd * box_losses[-1])

        return sum_losses, cls_losses, box_losses