def benchmark_region_loss():
"""
CommandLine:
python ~/code/netharn/netharn/models/yolo2/light_region_loss.py benchmark_region_loss --profile
Benchmark:
>>> benchmark_region_loss()
"""
from netharn.models.yolo2.light_yolo import Yolo
torch.random.manual_seed(0)
network = Yolo(num_classes=2, conf_thresh=4e-2)
self = light_region_loss.RegionLoss(num_classes=network.num_classes,
anchors=network.anchors)
Win, Hin = 96, 96
# true boxes for each item in the batch
# each box encodes class, center, width, and height
# coordinates are normalized in the range 0 to 1
# items in each batch are padded with dummy boxes with class_id=-1
target = torch.FloatTensor([
# boxes for batch item 1
[[0, 0.50, 0.50, 1.00, 1.00],
[1, 0.32, 0.42, 0.22, 0.12]],
# boxes for batch item 2 (it has no objects, note the pad!)
[[-1, 0, 0, 0, 0],
[-1, 0, 0, 0, 0]],
])
im_data = torch.randn(len(target), 3, Hin, Win)
output = network.forward(im_data)
import ubelt
for timer in ubelt.Timerit(250, bestof=10, label='time'):
with timer:
loss = float(self.forward(output, target))
print('loss = {!r}'.format(loss))
def test_tensor_brambox_compat_ground_truth_loss():
"""
Check that brambox and tensor boxes produce the same ground truth encoding
"""
network = Yolo(num_classes=2, conf_thresh=4e-2)
self = RegionLoss(num_classes=network.num_classes, anchors=network.anchors)
# Win, Hin = 96, 96
Win, Hin = 416, 416
target_brambox, target_tensor = demodata_targets(input_size=(Win, Hin))
# construct dummy data and run it through the network
im_data = torch.randn(len(target_brambox), 3, Hin, Win)
output = network.forward(im_data)
print('')
# Brambox and Tensor formats should produce the same loss
loss_brambox = self(output, target_brambox).data.cpu().numpy()
loss_tensor = self(output, target_tensor).data.cpu().numpy()
print('loss_brambox = {!r}'.format(loss_brambox))
print('loss_tensor = {!r}'.format(loss_tensor))
assert np.all(np.isclose(loss_brambox, loss_tensor))
# Brambox and Tensor formats should produce the same loss
loss_brambox = self(output, target_brambox, seen=9999).data.cpu().numpy()
loss_tensor = self(output, target_tensor, seen=9999).data.cpu().numpy()
print('loss_brambox = {!r}'.format(loss_brambox))
print('loss_tensor = {!r}'.format(loss_tensor))
assert np.all(np.isclose(loss_brambox, loss_tensor))
def test_tensor_brambox_compat_ground_truth_encoding():
"""
Check that brambox and tensor boxes produce the same ground truth encoding
"""
network = Yolo(num_classes=2, conf_thresh=4e-2)
self = RegionLoss(num_classes=network.num_classes, anchors=network.anchors)
Win, Hin = 96, 96
target_brambox, target_tensor = demodata_targets(input_size=(Win, Hin))
# convert network input size (e.g. 416x416) to output size (e.g. 13x13)
nW, nH = (Win // self.reduction, Hin // self.reduction)
pred_boxes = torch.rand(90, 4)
pred_confs = torch.rand(90)
_ret_ten = self._build_targets_tensor(pred_boxes, pred_confs,
target_tensor, nH, nW)
_ret_box = self._build_targets_brambox(pred_boxes, pred_confs,
target_brambox, nH, nW)
keys = ['coord_mask', 'conf_mask', 'cls_mask', 'tcoord', 'tconf', 'tcls']
errors = []
for key, item1, item2 in zip(keys, _ret_box, _ret_ten):
bad_flags = ~np.isclose(item1, item2)
if np.any(bad_flags):
bad_idxs = np.where(bad_flags)
msg = 'key={} did not agree at indices {}. values1={} values2={}'.format(
key, bad_idxs, item1[bad_flags], item2[bad_flags])
errors.append(msg)
if errors:
raise AssertionError('\n---\n'.join(errors))
def compare_loss_speed():
"""
python ~/code/netharn/netharn/models/yolo2/light_region_loss.py compare_loss_speed
Example:
>>> compare_loss_speed()
"""
from netharn.models.yolo2.light_yolo import Yolo
import netharn.models.yolo2.light_region_loss
import lightnet.network
import ubelt as ub
torch.random.manual_seed(0)
network = Yolo(num_classes=2, conf_thresh=4e-2)
self1 = netharn.models.yolo2.light_region_loss.RegionLoss(
num_classes=network.num_classes, anchors=network.anchors)
self2 = lightnet.network.RegionLoss(num_classes=network.num_classes,
anchors=network.anchors)
# Win, Hin = 416, 416
Win, Hin = 96, 96
# ----- More targets -----
rng = util.ensure_rng(0)
import netharn as nh
bsize = 4
# Make a random semi-realistic set of groundtruth items
n_targets = [rng.randint(0, 10) for _ in range(bsize)]
target_list = [
torch.FloatTensor(
np.hstack([
rng.randint(0, network.num_classes, nT)[:, None],
util.Boxes.random(nT, scale=1.0, rng=rng).data
])) for nT in n_targets
]
target = nh.data.collate.padded_collate(target_list)
im_data = torch.randn(len(target), 3, Hin, Win)
output = network.forward(im_data)
self1.iou_mode = 'c'
for timer in ub.Timerit(100, bestof=10, label='cython_ious'):
with timer:
loss_cy = float(self1(output, target))
self1.iou_mode = 'py'
for timer in ub.Timerit(100, bestof=10, label='python_ious'):
with timer:
loss_py = float(self1(output, target))
for timer in ub.Timerit(100, bestof=10, label='original'):
with timer:
loss_orig = float(self2(output, target))
print('loss_cy = {!r}'.format(loss_cy))
print('loss_py = {!r}'.format(loss_py))
print('loss_orig = {!r}'.format(loss_orig))
def profile_loss_speed():
"""
python ~/code/netharn/netharn/models/yolo2/light_region_loss.py profile_loss_speed --profile
Benchmark:
>>> profile_loss_speed()
"""
from netharn.models.yolo2.light_yolo import Yolo
import netharn.models.yolo2.light_region_loss
import lightnet.network
import netharn as nh
rng = util.ensure_rng(0)
torch.random.manual_seed(0)
network = Yolo(num_classes=2, conf_thresh=4e-2)
self1 = netharn.models.yolo2.light_region_loss.RegionLoss(
num_classes=network.num_classes, anchors=network.anchors)
self2 = lightnet.network.RegionLoss(num_classes=network.num_classes,
anchors=network.anchors)
bsize = 8
# Make a random semi-realistic set of groundtruth items
n_targets = [rng.randint(0, 20) for _ in range(bsize)]
target_list = [
torch.FloatTensor(
np.hstack([
rng.randint(0, network.num_classes, nT)[:, None],
util.Boxes.random(nT, scale=1.0, rng=rng).data
])) for nT in n_targets
]
target = nh.data.collate.padded_collate(target_list)
Win, Hin = 416, 416
im_data = torch.randn(len(target), 3, Hin, Win)
output = network.forward(im_data)
loss1 = float(self1(output, target))
loss2 = float(self2(output, target))
print('loss1 = {!r}'.format(loss1))
print('loss2 = {!r}'.format(loss2))