def track(self, model: model_interface.TrackerInterface, **kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
"""
super().track(model)
outputs = self._convert(model.get_output())
targets = self._convert(model.get_labels())
batch_idx = self._convert(model.get_batch())
if batch_idx is None:
raise ValueError(
"Your model need to set the batch_idx variable in its set_input function."
)
nb_batches = batch_idx.max() + 1
# pred to the groundtruth classes (selected by seg_classes[cat])
for b in range(nb_batches):
segl = targets[batch_idx == b]
cat = self._seg_to_class[segl[0]]
logits = outputs[batch_idx == b, :] # (num_points, num_classes)
segp = logits[:, self._class_seg_map[cat]].argmax(
1) + self._class_seg_map[cat][0]
part_ious = np.zeros(len(self._class_seg_map[cat]))
for l in self._class_seg_map[cat]:
if np.sum((segl == l) | (segp == l)) == 0:
# part is not present in this shape
part_ious[l - self._class_seg_map[cat][0]] = 1
else:
part_ious[l - self._class_seg_map[cat][0]] = float(
np.sum((segl == l) & (segp == l))) / float(
np.sum((segl == l) | (segp == l)))
self._shape_ious[cat].append(np.mean(part_ious))
self._miou_per_class, self._Cmiou, self._Imiou = self._get_metrics_per_class(
)
def track(self, model: model_interface.TrackerInterface, **kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
"""
if not self._dataset.has_labels(self._stage):
return
super().track(model)
outputs = model.get_output()
targets = model.get_labels()
# Mask ignored label
mask = targets != self._ignore_label
outputs = outputs[mask]
targets = targets[mask]
outputs = self._convert(outputs)
targets = self._convert(targets)
if len(targets) == 0:
return
assert outputs.shape[0] == len(targets)
self._confusion_matrix.count_predicted_batch(targets, np.argmax(outputs, 1))
self._acc = 100 * self._confusion_matrix.get_overall_accuracy()
self._macc = 100 * self._confusion_matrix.get_mean_class_accuracy()
self._miou = 100 * self._confusion_matrix.get_average_intersection_union()
def track(self, model: TrackerInterface, data=None, track_boxes=False, **kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
if tracking boxes, you must provide a labeled "data" object with the following attributes:
- id_scan: id of the scan to which the boxes belong to
- instance_box_cornerimport torchnet as tnts - gt box corners
- box_label_mask - mask for boxes (0 = no box)
- sem_cls_label - semantic label for each box
"""
super().track(model)
outputs: VoteNetResults = model.get_output()
total_num_proposal = outputs.objectness_label.shape[0] * outputs.objectness_label.shape[1]
pos_ratio = torch.sum(outputs.objectness_label.float()).item() / float(total_num_proposal)
self._pos_ratio.add(pos_ratio)
self._neg_ratio.add(torch.sum(outputs.objectness_mask.float()).item() / float(total_num_proposal) - pos_ratio)
obj_pred_val = torch.argmax(outputs.objectness_scores, 2) # B,K
self._obj_acc.add(
torch.sum((obj_pred_val == outputs.objectness_label.long()).float() * outputs.objectness_mask).item()
/ (torch.sum(outputs.objectness_mask) + 1e-6).item()
)
if data is None or self._stage == "train" or not track_boxes:
return
self._add_box_pred(outputs, data, model.conv_type)
def track(self, model: model_interface.TrackerInterface, **kwargs):
super().track(model)
if self._stage != "train":
batch_idx, batch_idx_target = model.get_batch()
batch_xyz, batch_xyz_target = model.get_xyz() # type: ignore
batch_ind, batch_ind_target, batch_size_ind = model.get_ind(
) # type: ignore
batch_feat, batch_feat_target = model.get_output()
nb_batches = batch_idx.max() + 1
cum_sum = 0
cum_sum_target = 0
begin = 0
end = batch_size_ind[0].item()
for b in range(nb_batches):
xyz = batch_xyz[batch_idx == b]
xyz_target = batch_xyz_target[batch_idx_target == b]
feat = batch_feat[batch_idx == b]
feat_target = batch_feat_target[batch_idx_target == b]
# as we have concatenated ind,
# we need to substract the cum_sum because we deal
# with each batch independently
# ind = batch_ind[b * len(batch_ind) / nb_batches : (b + 1) * len(batch_ind) / nb_batches] - cum_sum
# ind_target = (batch_ind_target[b * len(batch_ind_target) / nb_batches : (b + 1) * len(batch_ind_target) / nb_batches]- cum_sum_target)
ind = batch_ind[begin:end] - cum_sum
ind_target = batch_ind_target[begin:end] - cum_sum_target
# print(begin, end)
if b < nb_batches - 1:
begin = end
end = begin + batch_size_ind[b + 1].item()
cum_sum += len(xyz)
cum_sum_target += len(xyz_target)
rand = torch.randperm(len(feat))[:self.num_points]
rand_target = torch.randperm(
len(feat_target))[:self.num_points]
matches_gt = torch.stack([ind, ind_target]).transpose(0, 1)
# print(matches_gt.max(0), len(xyz), len(xyz_target), len(matches_gt))
# print(batch_ind.shape, nb_batches)
T_gt = estimate_transfo(xyz[matches_gt[:, 0]],
xyz_target[matches_gt[:, 1]])
matches_pred = get_matches(feat[rand],
feat_target[rand_target])
T_pred = fast_global_registration(
xyz[rand][matches_pred[:, 0]],
xyz_target[rand_target][matches_pred[:, 1]])
hit_ratio = compute_hit_ratio(
xyz[rand][matches_pred[:, 0]],
xyz_target[rand_target][matches_pred[:,
1]], T_gt, self.tau_1)
trans_error, rot_error = compute_transfo_error(T_pred, T_gt)
self._hit_ratio.add(hit_ratio.item())
self._feat_match_ratio.add(
float(hit_ratio.item() > self.tau_2))
self._trans_error.add(trans_error.item())
self._rot_error.add(rot_error.item())
def track(self,
model: model_interface.TrackerInterface,
full_res: bool = False,
data: Data = None,
**kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
"""
super().track(model)
self._conv_type = model.conv_type
outputs = self._convert(model.get_output())
targets = self._convert(model.get_labels())
batch_idx = self._convert(model.get_batch())
if batch_idx is None:
raise ValueError(
"Your model need to set the batch_idx variable in its set_input function."
)
nb_batches = batch_idx.max() + 1
if self._stage != "train" and full_res:
self._add_votes(data, outputs, batch_idx)
# pred to the groundtruth classes (selected by seg_classes[cat])
for b in range(nb_batches):
segl = targets[batch_idx == b]
cat = self._seg_to_class[segl[0]]
logits = outputs[batch_idx == b, :] # (num_points, num_classes)
segp = logits[:, self._class_seg_map[cat]].argmax(
1) + self._class_seg_map[cat][0]
part_ious = self._compute_part_ious(segl, segp, cat)
self._shape_ious[cat].append(np.mean(part_ious))
self._miou_per_class, self._Cmiou, self._Imiou = ShapenetPartTracker._get_metrics_per_class(
self._shape_ious)
def track(self, model: model_interface.TrackerInterface, **kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
"""
super().track(model)
outputs = model.get_output()
# targets = model.get_labels().flatten()
targets = model.get_labels()
avg_loss_dimensions, avg_loss_epsilons, avg_loss_offsets, mae_a1, mae_a2, mae_a3, mae_x0, mae_y0, mae_z0, mae_e1, mae_e2 = self.compute_loss_by_components(
outputs, targets)
self._loss_dimension.add(avg_loss_dimensions.detach().cpu().numpy())
self._loss_epsilon.add(avg_loss_epsilons.detach().cpu().numpy())
self._loss_offset.add(avg_loss_offsets.detach().cpu().numpy())
self._loss_mae_a1.add(mae_a1.detach().cpu().numpy())
self._loss_mae_a2.add(mae_a2.detach().cpu().numpy())
self._loss_mae_a3.add(mae_a3.detach().cpu().numpy())
self._loss_mae_x0.add(mae_x0.detach().cpu().numpy())
self._loss_mae_y0.add(mae_y0.detach().cpu().numpy())
self._loss_mae_z0.add(mae_z0.detach().cpu().numpy())
self._loss_mae_e1.add(mae_e1.detach().cpu().numpy())
self._loss_mae_e2.add(mae_e2.detach().cpu().numpy())
def track(self, model: model_interface.TrackerInterface, full_res=False, data=None, **kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
"""
super().track(model)
# Train mode or low res, nothing special to do
if self._stage == "train" or not full_res:
return
# Test mode, compute votes in order to get full res predictions
if self._test_area is None:
self._test_area = self._dataset.test_data.clone()
if self._test_area.y is None:
raise ValueError("It seems that the test area data does not have labels (attribute y).")
self._test_area.prediction_count = torch.zeros(self._test_area.y.shape[0], dtype=torch.int)
self._test_area.votes = torch.zeros((self._test_area.y.shape[0], self._num_classes), dtype=torch.float)
self._test_area.to(model.device)
# Gather origin ids and check that it fits with the test set
inputs = data if data is not None else model.get_input()
if inputs[SaveOriginalPosId.KEY] is None:
raise ValueError("The inputs given to the model do not have a %s attribute." % SaveOriginalPosId.KEY)
originids = inputs[SaveOriginalPosId.KEY]
if originids.dim() == 2:
originids = originids.flatten()
if originids.max() >= self._test_area.pos.shape[0]:
raise ValueError("Origin ids are larger than the number of points in the original point cloud.")
# Set predictions
outputs = model.get_output()
self._test_area.votes[originids] += outputs
self._test_area.prediction_count[originids] += 1
def track(self, model: model_interface.TrackerInterface, **kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
"""
super().track(model)
outputs = model.get_output()
targets = model.get_labels().flatten()
self._acc.add(100 * self.compute_acc(outputs, targets))
def track(self, model: model_interface.TrackerInterface, **kwargs):
""" Add model predictions (accuracy)
"""
super().track(model)
outputs = self._convert(model.get_output())
N = len(outputs) // 2
self._acc = compute_accuracy(outputs[:N], outputs[N:])
def track(self, model: TrackerInterface, **kwargs):
""" Track metrics for panoptic segmentation
"""
BaseTracker.track(self, model)
outputs: PanopticResults = model.get_output()
labels: PanopticLabels = model.get_labels()
# Track semantic
super()._compute_metrics(outputs.semantic_logits, labels.y)
def track(self, model: model_interface.TrackerInterface, **kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
"""
if not self._dataset.has_labels(self._stage):
return
super().track(model)
outputs = model.get_output()
targets = model.get_labels()
self._compute_metrics(outputs, targets)
def track(self, model: model_interface.TrackerInterface, full_res=False, **kwargs):
""" Add current model predictions (usually the result of a batch) to the tracking
"""
super().track(model)
# Set conv type
self._conv_type = model.conv_type
# Train mode or low res, nothing special to do
if not full_res or self._stage == "train" or kwargs.get("data") is None:
return
self._vote(kwargs.get("data"), model.get_output())
def track(self,
model: TrackerInterface,
data=None,
iou_threshold=0.25,
track_instances=True,
min_cluster_points=10,
**kwargs):
""" Track metrics for panoptic segmentation
"""
self._iou_threshold = iou_threshold
BaseTracker.track(self, model)
outputs: PanopticResults = model.get_output()
labels: PanopticLabels = model.get_labels()
# Track semantic
super()._compute_metrics(outputs.semantic_logits, labels.y)
if not data:
return
assert data.pos.dim() == 2, "Only supports packed batches"
# Object accuracy
clusters = PanopticTracker._extract_clusters(outputs,
min_cluster_points)
if not clusters:
return
predicted_labels = outputs.semantic_logits.max(1)[1]
tp, fp, acc = self._compute_acc(clusters, predicted_labels, labels,
data.batch, labels.num_instances,
iou_threshold)
self._pos.add(tp)
self._neg.add(fp)
self._acc_meter.add(acc)
# Track instances for AP
if track_instances:
pred_clusters = self._pred_instances_per_scan(
clusters, predicted_labels, outputs.cluster_scores, data.batch,
self._scan_id_offset)
gt_clusters = self._gt_instances_per_scan(labels.instance_labels,
labels.y, data.batch,
self._scan_id_offset)
self._ap_meter.add(pred_clusters, gt_clusters)
self._scan_id_offset += data.batch[-1].item() + 1
