def assign_primaries2(primaries, clusts, data):
"""
for each EM primary assign closest cluster that matches batch and group
data should contain groups of voxels
this version does not filter out compton clusters first
"""
primaries = primaries.cpu()
data = data.cpu()
labels = get_cluster_label(data, clusts)
batches = get_cluster_batch(data, clusts)
assn = []
for primary in primaries:
# get list of indices that match label and batch
pbatch = primary[-2]
# plabel = primary[-1]
# pselection = np.logical_and(labels == plabel, batches == pbatch)
pselection = batches == pbatch
pinds = np.where(pselection)[0] # indices to compare against
if len(pinds) < 1:
continue
scores = score_clusters_primary(clusts[pinds], data, labels[pinds],
primary)
ind = np.argmin(scores)
# print(scores[ind])
assn.append(pinds[ind])
return assn
def forward(self, data):
"""
Input:
data[0]: (Nx5) Cluster tensor with row (x, y, z, batch_id, cluster_id)
Output:
dictionary, with
'node_pred': torch.tensor with node prediction weights
"""
# Get device
cluster_label = data[0]
device = cluster_label.device
# Find index of points that belong to the same EM clusters
clusts = form_clusters_new(cluster_label)
# If requested, remove clusters below a certain size threshold
if self.remove_compton:
selection = np.where(filter_compton(clusts,
self.compton_thresh))[0]
if not len(selection):
return self.default_return(device)
clusts = clusts[selection]
# Get the cluster ids of each processed cluster
clust_ids = get_cluster_label(cluster_label, clusts)
# Get the batch ids of each cluster
batch_ids = get_cluster_batch(cluster_label, clusts)
# Form a complete graph (should add options for other structures, TODO)
edge_index = complete_graph(batch_ids, device=device)
if not edge_index.shape[0]:
return self.default_return(device)
# Obtain vertex features
x = cluster_vtx_features(cluster_label, clusts, device=device)
# Obtain edge features
e = cluster_edge_features(cluster_label,
clusts,
edge_index,
device=device)
# Convert the the batch IDs to a torch tensor to pass to Torch
xbatch = torch.tensor(batch_ids).to(device)
# Pass through the model, get output
out = self.node_predictor(x, edge_index, e, xbatch)
return {
**out, 'clust_ids': [torch.tensor(clust_ids)],
'batch_ids': [torch.tensor(batch_ids)],
'edge_index': [edge_index]
}
def assign_primaries(primaries,
clusts,
data,
use_labels=False,
max_dist=None,
compton_thresh=0):
"""
for each EM primary assign closest cluster that matches batch and group
data should contain groups of voxels
"""
primaries = primaries.cpu().detach().numpy()
data = data.cpu().detach().numpy()
#first remove compton-like clusters from list
selection = filter_compton(clusts,
compton_thresh) # non-compton looking clusters
selinds = np.where(selection)[0] # selected indices
cs2 = clusts[selinds]
# if everything looks compton, say no primaries
if len(cs2) < 1:
return []
if use_labels:
labels = get_cluster_label(data, cs2)
batches = get_cluster_batch(data, cs2)
assn = []
for primary in primaries:
# get list of indices that match label and batch
pbatch = primary[-2]
if use_labels:
plabel = primary[-1]
pselection = np.logical_and(labels == plabel, batches == pbatch)
else:
pselection = batches == pbatch
pinds = np.where(pselection)[0] # indices to compare against
if len(pinds) < 1:
continue
scores = score_clusters_primary(cs2[pinds], data, primary)
ind = np.argmin(scores)
if max_dist and scores[ind] > max_dist:
continue
# print(scores[ind])
assn.append(selinds[pinds[ind]])
# assignments may not be unique
assn = np.unique(assn)
return assn
def assign_primaries_unique(primaries, clusts, data, use_labels=False):
"""
for each EM primary assign closest cluster that matches batch and group
data should contain groups of voxels
"""
#first remove compton-like clusters from list
cs2 = clusts
# selection = filter_compton(clusts) # non-compton looking clusters
# selinds = np.where(selection)[0] # selected indices
# cs2 = clusts[selinds]
# if everything looks compton, say no primaries
if len(cs2) < 1:
return []
labels = get_cluster_label(data, cs2)
batches = get_cluster_batch(data, cs2)
assn = -1 * np.ones(len(primaries))
assn_scores = -1 * np.ones(len(primaries))
for i in range(len(primaries)):
primary = primaries[i]
# get list of indices that match label and batch
pbatch = primary[-2]
if use_labels:
plabel = primary[-1]
pselection = np.logical_and(labels == plabel, batches == pbatch)
else:
pselection = batches == pbatch
pinds = np.where(pselection)[0] # indices to compare against
if len(pinds) < 1:
continue
scores = score_clusters_primary(cs2[pinds], data, primary)
ind = np.argmin(scores)
pind = pinds[ind]
score = scores[ind]
already_assigned = np.where(assn == pind)[0]
if len(already_assigned) > 0:
current_low = assn_scores[already_assigned][0]
if score < current_low:
assn_scores[already_assigned] = -1.0
assn[already_assigned] = -1.0
else:
continue
assn_scores[i] = score
assn[i] = pind
return assn
def assign_primaries3(primaries, clusts, data):
"""
for each EM primary assign closest cluster that matches batch and group
data should contain groups of voxels
"""
#first remove compton-like clusters from list
cs2 = clusts
# selection = filter_compton(clusts) # non-compton looking clusters
# selinds = np.where(selection)[0] # selected indices
# cs2 = clusts[selinds]
# if everything looks compton, say no primaries
if len(cs2) < 1:
return []
labels = get_cluster_label(data, cs2)
batches = get_cluster_batch(data, cs2)
assn = []
for primary in primaries:
# get list of indices that match label and batch
pbatch = primary[-2]
plabel = primary[-1]
pselection = np.logical_and(labels == plabel, batches == pbatch)
pinds = np.where(pselection)[0] # indices to compare against
if len(pinds) < 1:
assn.append(-1)
continue
scores = score_clusters_primary(cs2[pinds], data, labels[pinds],
primary)
ind = np.argmin(scores)
# print(scores[ind])
# assn.append(selinds[pinds[ind]])
assn.append(pinds[ind])
return assn
def forward(self, out, clusters, groups, primary):
"""
out:
array output from the DataParallel gather function
out[0] - n_gpus tensors of edge indexes
out[1] - n_gpus tensors of predicted edge weights from model forward
out[2] - n_gpus arrays of group ids for each cluster
out[3] - n_gpus number of iterations
data:
cluster_labels - n_gpus Nx5 tensors of (x, y, z, batch_id, cluster_id)
group_labels - n_gpus Nx5 tensors of (x, y, z, batch_id, group_id)
em_primaries - n_gpus tensor of (x, y, z) coordinates of origins of EM primaries
"""
total_loss, total_acc, total_primary_fdr, total_primary_acc, total_iter = 0., 0., 0., 0., 0
ngpus = len(clusters)
for i in range(ngpus):
data0 = clusters[i]
data1 = groups[i]
data2 = primary[i]
clusts = form_clusters_new(data0)
# remove compton clusters
# if no cluster fits this condition, return
if self.remove_compton:
selection = filter_compton(
clusts) # non-compton looking clusters
if not len(selection):
edge_pred = out[1][i]
total_loss += self.lossfn(edge_pred, edge_pred)
total_acc += 1.
continue
clusts = clusts[selection]
# process group data
data_grp = data1
# form primary/secondary bipartite graph
primaries = assign_primaries(data2, clusts, data0)
batch = get_cluster_batch(data0, clusts)
# edge_index = primary_bipartite_incidence(batch, primaries)
group = get_cluster_label(data_grp, clusts)
primaries_true = assign_primaries(data2,
clusts,
data1,
use_labels=True)
primary_fdr, primary_tdr, primary_acc = analyze_primaries(
primaries, primaries_true)
total_primary_fdr += primary_fdr
total_primary_acc += primary_acc
niter = out[3][i][0] # number of iterations
total_iter += niter
for j in range(niter):
# determine true assignments
edge_index = out[0][i][j]
edge_assn = edge_assignment(edge_index,
batch,
group,
cuda=True)
edge_pred = out[1][i][j]
# print(edge_pred)
# print(edge_assn.shape)
# print(edge_pred.shape)
edge_assn = edge_assn.view(-1)
edge_pred = edge_pred.view(-1)
# print(edge_assn.shape)
# print(edge_pred.shape)
if self.balance:
edge_assn, edge_pred = self.balance_classes(
edge_assn, edge_pred)
total_loss += self.lossfn(edge_pred, edge_assn)
# compute accuracy of assignment
# need to multiply by batch size to be accurate
#total_acc = (np.max(batch) + 1) * torch.tensor(secondary_matching_vox_efficiency(edge_index, edge_assn, edge_pred, primaries, clusts, len(clusts)))
# use out['matched']
total_acc += torch.tensor(
secondary_matching_vox_efficiency2(out[2][i], group, primaries,
clusts))
return {
'primary_fdr': total_primary_fdr / ngpus,
'primary_acc': total_primary_acc / ngpus,
'accuracy': total_acc / ngpus,
'loss': total_loss / ngpus,
'n_iter': total_iter
}
def forward(self, out, clusters, groups, primary):
"""
out:
array output from the DataParallel gather function
out[0] - n_gpus tensors of edge indexes
out[1] - n_gpus tensors of predicted edge weights from model forward
out[2] - n_gpus arrays of group ids for each cluster
out[3] - n_gpus number of iterations
data:
cluster_labels - n_gpus Nx5 tensors of (x, y, z, batch_id, cluster_id)
group_labels - n_gpus Nx5 tensors of (x, y, z, batch_id, group_id)
em_primaries - n_gpus tensor of (x, y, z) coordinates of origins of EM primaries
"""
total_loss, total_acc, total_primary_fdr, total_primary_acc, total_iter = 0., 0., 0., 0., 0
total_ari, total_ami, total_sbd, total_pur, total_eff = 0., 0., 0., 0., 0.
ngpus = len(clusters)
for i in range(ngpus):
data0 = clusters[i]
data1 = groups[i]
data2 = primary[i]
clusts = form_clusters_new(data0)
# remove compton clusters
# if no cluster fits this condition, return
if self.remove_compton:
selection = filter_compton(
clusts) # non-compton looking clusters
if not len(selection):
edge_pred = out[1][i][0]
total_loss += self.lossfn(edge_pred, edge_pred)
total_acc += 1.
clusts = clusts[selection]
# process group data
data_grp = data1
# form primary/secondary bipartite graph
primaries = assign_primaries(data2, clusts, data0)
batch = get_cluster_batch(data0, clusts)
# edge_index = primary_bipartite_incidence(batch, primaries)
group = get_cluster_label(data_grp, clusts)
primaries_true = assign_primaries(data2,
clusts,
data1,
use_labels=True)
primary_fdr, primary_tdr, primary_acc = analyze_primaries(
primaries, primaries_true)
total_primary_fdr += primary_fdr
total_primary_acc += primary_acc
niter = out[3][i][0] # number of iterations
total_iter += niter
# loop over iterations and add loss at each iter.
for j in range(niter):
# determine true assignments
edge_index = out[0][i][j]
edge_assn = edge_assignment(edge_index,
batch,
group,
cuda=True,
dtype=torch.long)
# get edge predictions (2 channels)
edge_pred = out[1][i][j]
edge_assn = edge_assn.view(-1)
total_loss += self.lossfn(edge_pred, edge_assn)
# compute accuracy of assignment
total_acc += secondary_matching_vox_efficiency2(
out[2][i], group, primaries, clusts)
# get clustering metrics
#print(out[2][i].shape)
ari, ami, sbd, pur, eff = DBSCAN_cluster_metrics2(
out[2][i].cpu().numpy(), clusts, group)
total_ari += ari
total_ami += ami
total_sbd += sbd
total_pur += pur
total_eff += eff
return {
'primary_fdr': total_primary_fdr / ngpus,
'primary_acc': total_primary_acc / ngpus,
'ARI': ari / ngpus,
'AMI': ami / ngpus,
'SBD': sbd / ngpus,
'purity': pur / ngpus,
'efficiency': eff / ngpus,
'accuracy': total_acc / ngpus,
'loss': total_loss / ngpus,
'n_iter': total_iter
}
def forward(self, out, data0, data1):
"""
out:
dictionary output from GNN Model
keys:
'edge_pred': predicted edge weights from model forward
data:
data[0] - DBSCAN data
data[1] - groups data
"""
edge_pred = out[0][0]
data0 = data0[0]
data1 = data1[0]
device = data0.device
# first decide what true edges should be
# need to form graph, then pass through GNN
# clusts = form_clusters(data0)
clusts = form_clusters_new(data0)
# remove compton clusters
# if no cluster fits this condition, return
if self.remove_compton:
selection = filter_compton(
clusts, self.compton_thresh) # non-compton looking clusters
if not len(selection):
total_loss = self.lossfn(edge_pred, edge_pred)
return {'accuracy': 1., 'loss': total_loss}
clusts = clusts[selection]
# process group data
# data_grp = process_group_data(data1, data0)
data_grp = data1
# form graph
batch = get_cluster_batch(data0, clusts)
edge_index = complete_graph(batch, device=device)
if not edge_index.shape[0]:
total_loss = self.lossfn(edge_pred, edge_pred)
return {'accuracy': 0., 'loss': total_loss}
group = get_cluster_label(data_grp, clusts)
# determine true assignments
edge_assn = edge_assignment(edge_index,
batch,
group,
device=device,
dtype=torch.long)
edge_assn = edge_assn.view(-1)
# total loss on batch
total_loss = self.lossfn(edge_pred, edge_assn)
# compute assigned clusters
fe = edge_pred[1, :] - edge_pred[0, :]
cs = assign_clusters_UF(edge_index, fe, len(clusts), thresh=0.0)
ari, ami, sbd, pur, eff = DBSCAN_cluster_metrics2(cs, clusts, group)
edge_ct = edge_index.shape[1]
return {
'ARI': ari,
'AMI': ami,
'SBD': sbd,
'purity': pur,
'efficiency': eff,
'accuracy': ari,
'loss': total_loss,
'edge_count': edge_ct
}
def forward(self, edge_pred, data0, data1, data2):
"""
edge_pred:
predicted edge weights from model forward
data:
data[0] - 5 types data
data[1] - groups data
data[2] - primary data
"""
data0 = data0[0]
data1 = data1[0]
data2 = data2[0]
# first decide what true edges should be
# need to form graph, then pass through GNN
# clusts = form_clusters(data0)
clusts = form_clusters_new(data0)
# remove track-like particles
# types = get_cluster_label(data0, clusts)
# selection = types > 1 # 0 or 1 are track-like
# clusts = clusts[selection]
# remove compton clusters
# if no cluster fits this condition, return
selection = filter_compton(clusts) # non-compton looking clusters
if not len(selection):
total_loss = self.lossfn(edge_pred, edge_pred)
return {'accuracy': 1., 'loss_seg': total_loss}
clusts = clusts[selection]
# process group data
# data_grp = process_group_data(data1, data0)
data_grp = data1
# form primary/secondary bipartite graph
primaries = assign_primaries(data2, clusts, data0)
batch = get_cluster_batch(data0, clusts)
edge_index = primary_bipartite_incidence(batch, primaries)
group = get_cluster_label(data_grp, clusts)
primaries_true = assign_primaries(data2,
clusts,
data1,
use_labels=True)
print("primaries (est): ", primaries)
print("primaries (true): ", primaries_true)
# determine true assignments
edge_assn = edge_assignment(edge_index, batch, group, cuda=True)
edge_assn = edge_assn.view(-1)
edge_pred = edge_pred.view(-1)
if self.balance:
# weight edges so that 0/1 labels appear equally often
ind0 = edge_assn == 0
ind1 = edge_assn == 1
# number in each class
n0 = torch.sum(ind0).float()
n1 = torch.sum(ind1).float()
print("n0 = ", n0, " n1 = ", n1)
# weights to balance classes
w0 = n1 / (n0 + n1)
w1 = n0 / (n0 + n1)
print("w0 = ", w0, " w1 = ", w1)
edge_assn[ind0] = w0 * edge_assn[ind0]
edge_assn[ind1] = w1 * edge_assn[ind1]
edge_pred = edge_pred.clone()
edge_pred[ind0] = w0 * edge_pred[ind0]
edge_pred[ind1] = w1 * edge_pred[ind1]
total_loss = self.lossfn(edge_pred, edge_assn)
# compute accuracy of assignment
# need to multiply by batch size to be accurate
total_acc = (np.max(batch) + 1) * torch.tensor(
secondary_matching_vox_efficiency(edge_index, edge_assn, edge_pred,
primaries, clusts, len(clusts)))
return {'accuracy': total_acc, 'loss_seg': total_loss}