def prep_eval_gt_tracks_worker(args):
model, name, scene, logdir, split_on_no_edge = args
ofn = os.path.join(logdir, "tracks", os.path.basename(name))
graph, detection_weight_features, connection_batch = torch.load(
name + '-%s-eval_graph' % model.feature_name)
promote_graph(graph)
tracks, gt_graph_frames = ground_truth_tracks(scene.ground_truth(),
graph,
add_gt_class=True)
if split_on_no_edge:
tracks = split_track_on_missing_edge(tracks)
len_stats = defaultdict(int)
for tr in tracks:
for det in tr:
if hasattr(det, 'cls'):
cls = det.cls
gt_cls = det.gt_cls if hasattr(det, 'gt_cls') else None
det.__dict__ = {}
det.cls = cls
if gt_cls is not None:
det.gt_cls = gt_cls
else:
det.__dict__ = {}
len_stats[len(tr)] += 1
save_pickle(tracks, ofn)
save_pickle({'track_length': len_stats}, ofn + '-stats')
return ofn
def convert_ground_truth(self):
for cam in range(1, 9):
filename = self.path + '/ground_truth/camera%d.pck' % cam
if not os.path.exists(filename):
gt = defaultdict(list)
for det in tqdm(self.ground_truth_detections(cam),
'Converting ground truth for camera %d' % cam):
gt[det.frame].append(det)
save_pickle(gt, filename)
def make_duke_test_video():
cam = 2
seq = []
scene = Duke('/home/hakan/src/duke').scene(cam)
for frame_idx, frame, detections in video_detections(scene, 54373, 10):
fn = "test/data/duke_frame_%d_%.8d.jpg" % (cam, frame_idx)
imsave(frame, fn)
seq.append((frame_idx, fn.replace("test/", ""), detections))
save_pickle(seq, "test/data/duke_test_seq_cam2_10.pck")
gt = scene.ground_truth()
gt = {f: gt[f] for f, _, _ in seq}
save_pickle(gt, "test/data/duke_test_seq_cam2_10_gt.pck")
def prep_eval_tracks_worker(args):
model, name, device, logdir = args
ofn = os.path.join(logdir, "tracks", os.path.basename(name))
graph, detection_weight_features, connection_batch = torch.load(
name + '-%s-eval_graph' % model.feature_name)
promote_graph(graph)
detection_weight_features = detection_weight_features.to(device)
connection_batch = connection_batch.to(device)
tracks = lp_track(graph, connection_batch, detection_weight_features,
model)
for tr in tracks:
for det in tr:
if hasattr(det, 'cls'):
cls = det.cls
det.__dict__ = {}
det.cls = cls
else:
det.__dict__ = {}
save_pickle(tracks, ofn)
return ofn
def prep_minimal_graph_diffs(dataset,
model,
threads=None,
limit=None,
skipped_ggd_types=()):
trainval = {'train': [], 'eval': []}
final_trainval = {n: [] for n in trainval.keys()}
diff_lists = {}
jobs = []
os.makedirs(os.path.join(dataset.cachedir, "minimal_graph_diff"),
exist_ok=True)
for part in trainval.keys():
if limit is None:
entries = graph_names(dataset, part)
elif isinstance(limit, list):
entries = limit[part]
else:
entries = graph_names(dataset, part)
shuffle(entries)
entries = entries[:limit]
for fn, cam in entries:
bfn = os.path.join(dataset.cachedir, "minimal_graph_diff",
model.feature_name + '-' + os.path.basename(fn))
jobs.append((dataset, cam, part, model, fn, bfn))
final_trainval[part].append(bfn)
trainval_name = os.path.join(
dataset.cachedir, "minimal_graph_diff",
"%s_%s_trainval.json" % (dataset.name, model.feature_name))
skipped_ggd_types_name = os.path.join(
dataset.cachedir, "minimal_graph_diff",
"%s_%s_skipped_ggd_types.pck" % (dataset.name, model.feature_name))
if os.path.exists(trainval_name) and os.path.exists(
skipped_ggd_types_name):
current_trainval = load_json(trainval_name)
for part in trainval.keys():
if set(current_trainval[part]) != set(final_trainval[part]):
break
else:
if load_pickle(skipped_ggd_types_name) == skipped_ggd_types:
return
for part in trainval.keys():
dn = os.path.join(
dataset.cachedir, "minimal_graph_diff",
"%s_%s_%s_mmaps" % (dataset.name, model.feature_name, part))
if os.path.exists(dn):
rmtree(dn)
diff_lists[part] = GraphDiffList(dn, model)
save_pickle(skipped_ggd_types, skipped_ggd_types_name)
for part, base_bfn, bfns in parallel(prep_minimal_graph_diff_worker, jobs,
threads,
"Prepping minimal graph diffs"):
trainval[part].append(base_bfn)
save_json(trainval, trainval_name)
for bfn in bfns:
graphdiff = torch.load(bfn)
lst = diff_lists[part]
for gd in graphdiff:
if gd.name not in skipped_ggd_types:
lst.append(gd)