def forward(self, x):
x = self.fc0(x)
x = relu(x) # Should use tanh but it's not stable now.
x = self.fc1(x)
x = relu(x) # Should use tanh but it's not stable now.
x = self.fc2(x)
return x
def forward(self, old_x):
if self.stride == 1:
x_proj, x = self.channel_shuffle(old_x)
elif self.stride == 2:
x_proj = old_x
x = old_x
x_gap = x.mean(axis=2, keepdims=True).mean(axis=3, keepdims=True)
x_gap = self.reduce(x_gap)
x = self.wnet1(x, x_gap)
x = self.bn1(x)
x = F.relu(x)
x = self.wnet2(x, x_gap)
x = self.bn2(x)
x = self.wnet3(x, x_gap)
x = self.bn3(x)
x = F.relu(x)
if self.stride == 2:
x_proj = self.wnet_proj_1(x_proj, x_gap)
x_proj = self.bn_proj_1(x_proj)
x_proj = self.wnet_proj_2(x_proj, x_gap)
x_proj = self.bn_proj_2(x_proj)
x_proj = F.relu(x_proj)
return F.concat((x_proj, x), 1)
def forward(self, x):
x = self.fc1(x)
x = F.relu(x)
x = self.fc2(x)
x = F.relu(x)
x = self.fc3(x)
x = F.relu(x)
x = self.fc4(x)
prob = F.softmax(x)
return prob
def forward(self, x):
identity = x
x = self.conv1(x)
x = self.bn1(x)
x = F.relu(x)
x = self.conv2(x)
x = self.bn2(x)
identity = self.downsample(identity)
x += identity
x = F.relu(x)
return x
def get_stats(self, x):
activations, flops = 0, 0
identity = x
in_x = deepcopy(x)
x = self.conv1(x)
tmp_flops, tmp_acts = cal_conv_stats(self.conv1, in_x, x)
activations += tmp_acts
flops += tmp_flops
in_x = deepcopy(x)
x = self.bn1(x)
tmp_flops, tmp_acts = cal_norm_stats(self.bn1, in_x, x)
activations += tmp_acts
flops += tmp_flops
x = F.relu(x)
in_x = deepcopy(x)
x = self.conv2(x)
tmp_flops, tmp_acts = cal_conv_stats(self.conv2, in_x, x)
activations += tmp_acts
flops += tmp_flops
in_x = deepcopy(x)
x = self.bn2(x)
tmp_flops, tmp_acts = cal_norm_stats(self.bn2, in_x, x)
activations += tmp_acts
flops += tmp_flops
if self.tmp_in_channels == self.tmp_channels and self.stride == 1:
identity = self.downsample_id(identity)
else:
in_x = deepcopy(identity)
identity = self.downsample_conv(identity)
tmp_flops, tmp_acts = cal_conv_stats(self.downsample_conv, in_x,
identity)
activations += tmp_acts
flops += tmp_flops
in_x = deepcopy(identity)
identity = self.downsample_norm(identity)
tmp_flops, tmp_acts = cal_norm_stats(self.downsample_norm, in_x,
identity)
activations += tmp_acts
flops += tmp_flops
x += identity
x = F.relu(x)
return x, flops, activations
def two_layer_conv(x):
# (8, 3, 3, 3) 代表（输出信道数，输入信道数，卷积核高度，卷积核宽度）
conv_weight = mge.Parameter(np.random.randn(8, 3, 3, 3).astype(np.float32))
# 对于 8 个卷积核，提供 8 个 bias
conv_bias = mge.Parameter(np.zeros((1, 8, 1, 1), dtype=np.float32))
x = F.conv2d(x, conv_weight, conv_bias)
x = F.relu(x)
conv_weight = mge.Parameter(
np.random.randn(16, 8, 3, 3).astype(np.float32))
conv_bias = mge.Parameter(np.zeros((1, 16, 1, 1), dtype=np.float32))
x = F.conv2d(x, conv_weight, conv_bias)
x = F.relu(x)
return x
def forward(self, x):
x1 = self.conv(x)
x1 = self.bn(x1)
x1 = F.relu(x1)
x1 = x1 * self.scale[0]
x2 = self.conv2(x)
x2 = self.bn2(x2)
x2 = F.relu(x2)
x2 = x2 * self.scale[1]
y = x1 + x2
y = y + 4
y = self.scale[0] + y
y = F.relu(y) * 3
return y
def forward(self, fpn_fms, rcnn_rois, im_info=None, gt_boxes=None):
rcnn_rois, labels, bbox_targets = self.get_ground_truth(
rcnn_rois, im_info, gt_boxes)
fpn_fms = [fpn_fms[x] for x in self.in_features]
pool_features = layers.roi_pool(
fpn_fms,
rcnn_rois,
self.stride,
self.pooling_size,
self.pooling_method,
)
flatten_feature = F.flatten(pool_features, start_axis=1)
roi_feature = F.relu(self.fc1(flatten_feature))
roi_feature = F.relu(self.fc2(roi_feature))
pred_cls = self.pred_cls(roi_feature)
pred_delta = self.pred_delta(roi_feature)
if self.training:
# loss for classification
loss_rcnn_cls = layers.softmax_loss(pred_cls, labels)
# loss for regression
pred_delta = pred_delta.reshape(-1, self.cfg.num_classes + 1, 4)
vlabels = labels.reshape(-1, 1).broadcast((labels.shapeof(0), 4))
pred_delta = F.indexing_one_hot(pred_delta, vlabels, axis=1)
loss_rcnn_loc = layers.get_smooth_l1_loss(
pred_delta,
bbox_targets,
labels,
self.cfg.rcnn_smooth_l1_beta,
norm_type="all",
)
loss_dict = {
'loss_rcnn_cls': loss_rcnn_cls,
'loss_rcnn_loc': loss_rcnn_loc
}
return loss_dict
else:
# slice 1 for removing background
pred_scores = F.softmax(pred_cls, axis=1)[:, 1:]
pred_delta = pred_delta[:, 4:].reshape(-1, 4)
target_shape = (rcnn_rois.shapeof(0), self.cfg.num_classes, 4)
# rois (N, 4) -> (N, 1, 4) -> (N, 80, 4) -> (N * 80, 4)
base_rois = F.add_axis(rcnn_rois[:, 1:5],
1).broadcast(target_shape).reshape(-1, 4)
pred_bbox = self.box_coder.decode(base_rois, pred_delta)
return pred_bbox, pred_scores
def forward(self, x):
identity = x
x = self.conv1(x)
x = self.bn1(x)
x = F.relu(x)
x = self.conv2(x)
x = self.bn2(x)
if self.tmp_in_channels == self.tmp_channels and self.stride == 1:
identity = self.downsample_id(identity)
else:
identity = self.downsample_conv(identity)
identity = self.downsample_norm(identity)
x += identity
x = F.relu(x)
return x
def forward(self, x):
"""
Residual block feedforward function.
"""
# NOTE: to completely reproduce pytorch, we use F.relu(x) to replace x in shortcut
# since pytorch use inplace relu in residual branch.
return self._residual(x) + self._shortcut(F.relu(x))
def forward(self, x):
x = self.conv0(x)
if self.bn0:
x = self.bn0(x)
x = F.relu(x)
x = self.pool0(x)
x = self.conv1(x)
if self.bn1:
x = self.bn1(x)
x = F.relu(x)
x = self.pool1(x)
x = F.flatten(x, 1)
x = self.fc0(x)
x = F.relu(x)
x = self.fc1(x)
return x
def forward(self, x):
x = self.t1(x)
x = F.relu(x)
x = self.t2(x)
x = F.relu(x)
x = self.t3(x)
x = F.relu(x)
x = self.t4(x)
x = F.relu(x)
logit = self.fc(x)
logit = logit.reshape(-1, 28)
#print('logit', logit.shape)
#print('logit0', logit[0])
#print('logit1', logit[1])
return logit
def test_replace_var():
a = Tensor([1, 2])
b = Tensor([3, 4])
@trace(symbolic=True, capture_as_const=True)
def fwd(a, b):
return (a + b) * 2
fwd(a, b)
orig_model = io.BytesIO()
fwd.dump(orig_model,
arg_names=["a", "b"],
output_names="o",
optimize_for_inference=False)
orig_model.seek(0)
graph = Net.load(orig_model)
vara = graph.var_filter.name("a").as_unique()
varb = graph.var_filter.name("b").as_unique()
out = F.mul(vara, varb)
out = F.relu(out)
opnode = list(graph.opr_filter.has_input(vara))
repl_dict = {opnode[0].outputs[0]: out}
graph.replace_vars(repl_dict)
modified_model = io.BytesIO()
graph.dump(modified_model)
modified_model.seek(0)
load_graph = GraphInference(modified_model)
out = load_graph.run(a, b)
np.testing.assert_equal(out["o"], [6, 16])
def test_replace_opr():
a = Tensor([1, 2])
b = Tensor([3, 4])
@trace(symbolic=True, capture_as_const=True)
def fwd(a, b):
return (a + b) * 2
fwd(a, b)
orig_model = io.BytesIO()
fwd.dump(orig_model,
arg_names=["a", "b"],
output_names="o",
optimize_for_inference=False)
orig_model.seek(0)
graph = Net.load(orig_model)
vara = graph.var_filter.name("a").as_unique()
varb = graph.var_filter.name("b").as_unique()
out1 = F.sub(vara, varb)
out1 = F.relu(out1)
out1 = graph.add_dep_oprs(out1)
orig_opr = graph.opr_filter.has_input(vara).as_unique()
repl_dict = {orig_opr: out1[0].owner}
graph.replace_oprs(repl_dict)
modified_model1 = io.BytesIO()
graph.dump(modified_model1)
modified_model1.seek(0)
load_graph = GraphInference(modified_model1)
out = load_graph.run(a, b)
np.testing.assert_equal(out["o"], [0, 0])
def forward(self, x):
# x: bxlxi
x = self.init_map(x) # bxlxe
ori = x
p = self.position_encoding(eye)
x = x + p
values = self.value_mapping(x) #bxlxe
keys = self.key_mapping(x) #bxlxe
querys = self.key_mapping(x)
#print('transformer', values.shape, keys.shape, querys.shape)
attention = F.softmax(F.batched_matrix_mul(querys,
keys.dimshuffle(0, 2, 1)),
axis=1) #bxlxl
#print(attention[0])
#print(attention[0].sum(axis=0))
#print('attention', attention.shape)
out = F.batched_matrix_mul(values.dimshuffle(0, 2, 1), attention)
out = out.dimshuffle(0, 2, 1)
out = out + ori
out = F.relu(out)
#a,b,c = out.shape[0], out.shape[1], out.shape[2]
#tmp = out.reshape(-1, self.key_embedding)
#i = tmp.shape[0]
#out = self.norm(tmp)
#out = out.reshape(a,b,c)
return out
def forward(self, x):
features = self.features(x)
out = F.relu(features)
out = F.avg_pool2d(out, kernel_size=7,
stride=1).reshape(features.shape[0], -1)
out = self.classifier(out)
return out
def forward(self, fpn_fms, rcnn_rois, im_info=None, gt_boxes=None):
rcnn_rois, labels, bbox_targets = self.get_ground_truth(
rcnn_rois, im_info, gt_boxes)
fpn_fms = [fpn_fms[x] for x in self.in_features]
pool_features = layers.roi_pool(
fpn_fms,
rcnn_rois,
self.stride,
self.pooling_size,
self.pooling_method,
)
flatten_feature = F.flatten(pool_features, start_axis=1)
roi_feature = F.relu(self.fc1(flatten_feature))
roi_feature = F.relu(self.fc2(roi_feature))
pred_logits = self.pred_cls(roi_feature)
pred_offsets = self.pred_delta(roi_feature)
if self.training:
# loss for rcnn classification
loss_rcnn_cls = F.loss.cross_entropy(pred_logits, labels, axis=1)
# loss for rcnn regression
pred_offsets = pred_offsets.reshape(-1, self.cfg.num_classes, 4)
num_samples = labels.shape[0]
fg_mask = labels > 0
loss_rcnn_bbox = layers.smooth_l1_loss(
pred_offsets[fg_mask, labels[fg_mask] - 1],
bbox_targets[fg_mask],
self.cfg.rcnn_smooth_l1_beta,
).sum() / F.maximum(num_samples, 1)
loss_dict = {
"loss_rcnn_cls": loss_rcnn_cls,
"loss_rcnn_bbox": loss_rcnn_bbox,
}
return loss_dict
else:
# slice 1 for removing background
pred_scores = F.softmax(pred_logits, axis=1)[:, 1:]
pred_offsets = pred_offsets.reshape(-1, 4)
target_shape = (rcnn_rois.shape[0], self.cfg.num_classes, 4)
# rois (N, 4) -> (N, 1, 4) -> (N, 80, 4) -> (N * 80, 4)
base_rois = F.broadcast_to(
F.expand_dims(rcnn_rois[:, 1:5], axis=1),
target_shape).reshape(-1, 4)
pred_bbox = self.box_coder.decode(base_rois, pred_offsets)
return pred_bbox, pred_scores
def forward(self, features, im_info, boxes=None):
# prediction
features = [features[x] for x in self.in_features]
# get anchors
all_anchors_list = [
self.anchors_generator(fm, stride)
for fm, stride in zip(features, self.stride_list)
]
pred_cls_logit_list = []
pred_bbox_offset_list = []
for x in features:
t = F.relu(self.rpn_conv(x))
scores = self.rpn_cls_score(t)
pred_cls_logit_list.append(
scores.reshape(
scores.shape[0],
2,
self.num_cell_anchors,
scores.shape[2],
scores.shape[3],
))
bbox_offsets = self.rpn_bbox_offsets(t)
pred_bbox_offset_list.append(
bbox_offsets.reshape(
bbox_offsets.shape[0],
self.num_cell_anchors,
4,
bbox_offsets.shape[2],
bbox_offsets.shape[3],
))
# sample from the predictions
rpn_rois = self.find_top_rpn_proposals(pred_bbox_offset_list,
pred_cls_logit_list,
all_anchors_list, im_info)
if self.training:
rpn_labels, rpn_bbox_targets = self.get_ground_truth(
boxes, im_info, all_anchors_list)
pred_cls_logits, pred_bbox_offsets = self.merge_rpn_score_box(
pred_cls_logit_list, pred_bbox_offset_list)
# rpn loss
loss_rpn_cls = layers.softmax_loss(pred_cls_logits, rpn_labels)
loss_rpn_loc = layers.get_smooth_l1_loss(
pred_bbox_offsets,
rpn_bbox_targets,
rpn_labels,
self.cfg.rpn_smooth_l1_beta,
norm_type="all",
)
loss_dict = {
"loss_rpn_cls": loss_rpn_cls,
"loss_rpn_loc": loss_rpn_loc
}
return rpn_rois, loss_dict
else:
return rpn_rois
def forward(self, x):
out = F.relu(self.bn1(self.conv1(x)))
out = self.layer1(out)
out = self.layer2(out)
out = self.layer3(out)
out = out.mean(3).mean(2)
out = self.linear(out)
return out
def forward(self, x):
identity = x
x = self.conv_bn_relu1(x)
x = self.conv_bn2(x)
identity = self.downsample(identity)
x = self.add(x, identity)
x = F.relu(x)
return x
def forward(self, features: List[Tensor]):
logits, offsets, ctrness = [], [], []
for feature, scale, stride in zip(features, self.scale_list, self.stride_list):
logits.append(self.cls_score(self.cls_subnet(feature)))
bbox_subnet = self.bbox_subnet(feature)
offsets.append(F.relu(self.bbox_pred(bbox_subnet) * scale) * stride)
ctrness.append(self.ctrness(bbox_subnet))
return logits, offsets, ctrness
def func(x, y):
a = x + y
a1 = F.relu(a)
a2 = F.abs(a)
a3 = F.ceil(a) * 2
a4 = F.floor(a)
r = a1 - a2
r1 = a3 / a4
return r, r1
def forward(self, x):
x = self.conv1(x)
x = self.bn1(x)
x = F.relu(x)
x = self.avgpool(x)
x = F.avg_pool2d(x, 22)
x = F.flatten(x, 1)
x = self.fc(x)
return x
def forward(self, fpn_fms, rcnn_rois, labels=None, bbox_targets=None):
# stride: 64,32,16,8,4 -> 4, 8, 16, 32
fpn_fms = fpn_fms[1:][::-1]
stride = [4, 8, 16, 32]
pool_features, rcnn_rois, labels, bbox_targets = roi_pool(
fpn_fms, rcnn_rois, stride, (7, 7), 'roi_align', labels,
bbox_targets)
flatten_feature = F.flatten(pool_features, start_axis=1)
roi_feature = F.relu(self.fc1(flatten_feature))
roi_feature = F.relu(self.fc2(roi_feature))
pred_cls = self.pred_cls(roi_feature)
pred_delta = self.pred_delta(roi_feature)
if self.training:
# loss for regression
labels = labels.astype(np.int32).reshape(-1)
# mulitple class to one
pos_masks = labels > 0
pred_delta = pred_delta.reshape(-1, config.num_classes, 4)
indexing_label = (labels * pos_masks).reshape(-1, 1)
indexing_label = indexing_label.broadcast((labels.shapeof()[0], 4))
pred_delta = F.indexing_one_hot(pred_delta, indexing_label, 1)
localization_loss = smooth_l1_loss(pred_delta, bbox_targets,
config.rcnn_smooth_l1_beta)
localization_loss = localization_loss * pos_masks
# loss for classification
valid_masks = labels >= 0
objectness_loss = softmax_loss(pred_cls, labels)
objectness_loss = objectness_loss * valid_masks
normalizer = 1.0 / (valid_masks.sum())
loss_rcnn_cls = objectness_loss.sum() * normalizer
loss_rcnn_loc = localization_loss.sum() * normalizer
loss_dict = {}
loss_dict['loss_rcnn_cls'] = loss_rcnn_cls
loss_dict['loss_rcnn_loc'] = loss_rcnn_loc
return loss_dict
else:
pred_scores = F.softmax(pred_cls)[:, 1:].reshape(-1, 1)
pred_delta = pred_delta[:, 4:].reshape(-1, 4)
target_shape = (rcnn_rois.shapeof()[0], config.num_classes - 1, 4)
base_rois = F.add_axis(rcnn_rois[:, 1:5],
1).broadcast(target_shape).reshape(-1, 4)
pred_bbox = restore_bbox(base_rois, pred_delta, True)
pred_bbox = F.concat([pred_bbox, pred_scores], axis=1)
return pred_bbox
def forward(self, fpn_fms, rcnn_rois, labels=None, bbox_targets=None):
# stride: 64,32,16,8,4 -> 4, 8, 16, 32
fpn_fms = fpn_fms[1:][::-1]
stride = [4, 8, 16, 32]
pool_features, rcnn_rois, labels, bbox_targets = roi_pool(
fpn_fms, rcnn_rois, stride, (7, 7), 'roi_align',
labels, bbox_targets)
flatten_feature = F.flatten(pool_features, start_axis=1)
roi_feature = F.relu(self.fc1(flatten_feature))
roi_feature = F.relu(self.fc2(roi_feature))
pred_emd_pred_cls_0 = self.emd_pred_cls_0(roi_feature)
pred_emd_pred_delta_0 = self.emd_pred_delta_0(roi_feature)
pred_emd_pred_cls_1 = self.emd_pred_cls_1(roi_feature)
pred_emd_pred_delta_1 = self.emd_pred_delta_1(roi_feature)
if self.training:
loss0 = emd_loss(
pred_emd_pred_delta_0, pred_emd_pred_cls_0,
pred_emd_pred_delta_1, pred_emd_pred_cls_1,
bbox_targets, labels)
loss1 = emd_loss(
pred_emd_pred_delta_1, pred_emd_pred_cls_1,
pred_emd_pred_delta_0, pred_emd_pred_cls_0,
bbox_targets, labels)
loss = F.concat([loss0, loss1], axis=1)
indices = F.argmin(loss, axis=1)
loss_emd = F.indexing_one_hot(loss, indices, 1)
loss_emd = loss_emd.sum()/loss_emd.shapeof()[0]
loss_dict = {}
loss_dict['loss_rcnn_emd'] = loss_emd
return loss_dict
else:
pred_scores_0 = F.softmax(pred_emd_pred_cls_0)[:, 1:].reshape(-1, 1)
pred_scores_1 = F.softmax(pred_emd_pred_cls_1)[:, 1:].reshape(-1, 1)
pred_delta_0 = pred_emd_pred_delta_0[:, 4:].reshape(-1, 4)
pred_delta_1 = pred_emd_pred_delta_1[:, 4:].reshape(-1, 4)
target_shape = (rcnn_rois.shapeof()[0], config.num_classes - 1, 4)
base_rois = F.add_axis(rcnn_rois[:, 1:5], 1).broadcast(target_shape).reshape(-1, 4)
pred_bbox_0 = restore_bbox(base_rois, pred_delta_0, True)
pred_bbox_1 = restore_bbox(base_rois, pred_delta_1, True)
pred_bbox_0 = F.concat([pred_bbox_0, pred_scores_0], axis=1)
pred_bbox_1 = F.concat([pred_bbox_1, pred_scores_1], axis=1)
#[{head0, pre1, tag1}, {head1, pre1, tag1}, {head0, pre1, tag2}, ...]
pred_bbox = F.concat((pred_bbox_0, pred_bbox_1), axis=1).reshape(-1,5)
return pred_bbox
def forward(self, fpn_fms, rcnn_rois, labels=None, bbox_targets=None):
# stride: 64,32,16,8,4 -> 4, 8, 16, 32
fpn_fms = fpn_fms[1:]
fpn_fms.reverse()
stride = [4, 8, 16, 32]
poo5, rcnn_rois, labels, bbox_targets = roi_pool(
fpn_fms, rcnn_rois, stride, (7, 7), 'roi_align',
labels, bbox_targets)
poo5 = F.flatten(poo5, start_axis=1)
fc1 = F.relu(self.fc1(poo5))
fc2 = F.relu(self.fc2(fc1))
cls_scores = self.cls(fc2)
pred_boxes = self.bbox(fc2)
# a = self.a(fc2)
# b = self.b(fc2)
# prob = F.stack([a, b], axis=1).reshape(-1, a.shape[1])
prob = F.concat([pred_boxes, cls_scores], axis=1)
if self.training:
# emd_loss = self.compute_gemini_loss(prob, bbox_targets, labels)
bbox_targets, labels = bbox_targets.reshape(-1, 4), labels.flatten()
cls_loss = softmax_loss(cls_scores, labels)
pred_boxes = pred_boxes.reshape(-1, self.n, 4)
bbox_loss = smooth_l1_loss_rcnn(pred_boxes, bbox_targets, labels, \
config.rcnn_smooth_l1_beta)
loss_dict = {}
loss_dict['cls_loss'] = cls_loss
loss_dict['bbox_loss'] = bbox_loss
return loss_dict
else:
offsets, cls_scores = prob[:, :-self.n], prob[:, -self.n:]
pred_bbox = offsets.reshape(-1, self.n, 4)
cls_prob = F.softmax(cls_scores, axis=1)
n = rcnn_rois.shape[0]
rois = F.broadcast_to(F.expand_dims(rcnn_rois[:, 1:5], axis=1), (n, 1, 4)).reshape(-1, 4)
normalized = config.rcnn_bbox_normalize_targets
pred_boxes = restore_bbox(rois, pred_bbox, normalized, config)
pred_bbox = F.concat([pred_boxes, F.expand_dims(cls_prob, axis=2)], axis=2)
return pred_bbox
def forward(self, fpn_fms, proposals, labels=None, bbox_targets=None):
# input p2-p5
fpn_fms = fpn_fms[1:][::-1]
stride = [4, 8, 16, 32]
#pool_features = roi_pooler(fpn_fms, proposals, stride, (7, 7), "ROIAlignV2")
pool_features, proposals, labels, bbox_targets = roi_pool(
fpn_fms, proposals, stride, (7, 7), 'roi_align',
labels, bbox_targets)
flatten_feature = F.flatten(pool_features, start_axis=1)
roi_feature = F.relu(self.fc1(flatten_feature))
roi_feature = F.relu(self.fc2(roi_feature))
pred_cls = self.pred_cls(roi_feature)
pred_delta = self.pred_delta(roi_feature)
if self.training:
# loss for regression
labels = labels.astype(np.int32).reshape(-1)
# mulitple class to one
pos_masks = labels > 0
localization_loss = smooth_l1_loss(
pred_delta,
bbox_targets,
config.rcnn_smooth_l1_beta)
localization_loss = localization_loss * pos_masks
# loss for classification
valid_masks = labels >= 0
objectness_loss = softmax_loss(
pred_cls,
labels)
objectness_loss = objectness_loss * valid_masks
normalizer = 1.0 / (valid_masks.sum())
loss_rcnn_cls = objectness_loss.sum() * normalizer
loss_rcnn_loc = localization_loss.sum() * normalizer
loss_dict = {}
loss_dict[self.stage_name + '_cls'] = loss_rcnn_cls
loss_dict[self.stage_name + '_loc'] = loss_rcnn_loc
pred_bbox = restore_bbox(proposals[:, 1:5], pred_delta, True)
pred_proposals = F.zero_grad(F.concat([proposals[:, 0].reshape(-1, 1), pred_bbox], axis=1))
return pred_proposals, loss_dict
else:
pred_scores = F.softmax(pred_cls)[:, 1].reshape(-1, 1)
pred_bbox = restore_bbox(proposals[:, 1:5], pred_delta, True)
pred_proposals = F.concat([proposals[:, 0].reshape(-1, 1), pred_bbox], axis=1)
return pred_proposals, pred_scores
def forward(self, fpn_fms, rcnn_rois, labels=None, bbox_targets=None):
# stride: 64,32,16,8,4 -> 4, 8, 16, 32
fpn_fms = fpn_fms[1:]
fpn_fms.reverse()
stride = [4, 8, 16, 32]
poo5, rcnn_rois, labels, bbox_targets = roi_pool(
fpn_fms, rcnn_rois, stride, (7, 7), 'roi_align', labels,
bbox_targets)
poo5 = F.flatten(poo5, start_axis=1)
fc1 = F.relu(self.fc1(poo5))
fc2 = F.relu(self.fc2(fc1))
a = self.a(fc2)
b = self.b(fc2)
prob = F.stack([a, b], axis=1).reshape(-1, a.shape[1])
if self.refinement:
final_prob = self.refinement_module(prob, fc2)
if self.training:
emd_loss = self.compute_gemini_loss(prob, bbox_targets, labels)
loss_dict = {}
loss_dict['loss_rcnn_emd'] = emd_loss
if self.refinement_module:
final_emd_loss = self.compute_gemini_loss(
final_prob, bbox_targets, labels)
loss_dict['final_rcnn_emd'] = final_emd_loss
return loss_dict
else:
offsets, cls_scores = prob[:, :-self.n], prob[:, -self.n:]
pred_bbox = offsets.reshape(-1, self.n, 4)
cls_prob = F.softmax(cls_scores, axis=1)
n = rcnn_rois.shape[0]
rois = F.broadcast_to(F.expand_dims(rcnn_rois[:, 1:5], axis=1),
(n, 2, 4)).reshape(-1, 4)
normalized = config.rcnn_bbox_normalize_targets
pred_boxes = restore_bbox(rois, pred_bbox, normalized, config)
pred_bbox = F.concat(
[pred_boxes, F.expand_dims(cls_prob, axis=2)], axis=2)
return pred_bbox
def run_conv2d(inp, w, b):
O = F.conv2d(
inp,
w,
b if has_bias else None,
stride=(SH, SW),
padding=(PH, PW),
)
if nonlinear_mode == "relu":
return F.relu(O)
else:
return O
def forward(self, fpn_fms, rcnn_rois, labels=None, bbox_targets=None):
# stride: 64,32,16,8,4 -> 4, 8, 16, 32
fpn_fms = fpn_fms[1:][::-1]
stride = [4, 8, 16, 32]
pool_features, rcnn_rois, labels, bbox_targets = roi_pool(
fpn_fms, rcnn_rois, stride, (7, 7), 'roi_align',
labels, bbox_targets)
flatten_feature = F.flatten(pool_features, start_axis=1)
roi_feature = F.relu(self.fc1(flatten_feature))
roi_feature = F.relu(self.fc2(roi_feature))
pred_cls = self.pred_cls(roi_feature)
pred_delta = self.pred_delta(roi_feature)
if self.training:
# loss for regression
labels = labels.astype(np.int32).reshape(-1)
pos_masks = labels > 0
localization_loss = smooth_l1_loss(
pred_delta,
bbox_targets,
config.rcnn_smooth_l1_beta)
localization_loss = localization_loss * pos_masks
# loss for classification
valid_masks = labels >= 0
objectness_loss = softmax_loss(
pred_cls,
labels)
objectness_loss = objectness_loss * valid_masks
normalizer = 1.0 / (valid_masks.sum())
loss_rcnn_cls = objectness_loss.sum() * normalizer
loss_rcnn_loc = localization_loss.sum() * normalizer
loss_dict = {}
loss_dict['loss_rcnn_cls'] = loss_rcnn_cls
loss_dict['loss_rcnn_loc'] = loss_rcnn_loc
return loss_dict
else:
pred_scores = F.softmax(pred_cls)
pred_bbox = restore_bbox(rcnn_rois[:, 1:5], pred_delta, True)
pred_bbox = F.concat([pred_bbox, pred_scores[:, 1].reshape(-1,1)], axis=1)
return pred_bbox
