def to_array(obj, copy_array=False):
if isinstance(obj, torch.Tensor):
if len(obj.shape) == 1 or obj.shape[1] == 1:
if copy_array:
return torch.clone(obj)
return obj
return torch.ravel(torch.FloatTensor(np.array([v for v in obj])))
def validation_epoch_end(self, outputs: Dict[str, Any]):
prediction = torch.cat([tmp["prediction"] for tmp in outputs])
target = torch.cat([tmp["target"] for tmp in outputs])
# --- masked cm, flatten all tensors, subset by forest pixels
lu = torch.ravel(torch.cat([tmp["lu"] for tmp in outputs]))
prediction_masked = torch.ravel(prediction)[lu == 1]
target_masked = torch.ravel(target)[lu == 1]
confusion_matrix = torchmetrics.functional.confusion_matrix(
prediction, target, normalize="true", num_classes=len(self.classes)
)
confusion_matrix_masked = torchmetrics.functional.confusion_matrix(
prediction_masked,
target_masked,
normalize="true",
num_classes=len(self.classes),
)
dfs = {}
for label, cm in zip(
["cm_norm", "cm_norm_masked"], [confusion_matrix, confusion_matrix_masked]
):
dfs[label] = pd.DataFrame(
cm.detach().cpu().numpy(),
index=self.classes,
columns=self.classes,
)
cm_chart = show_cm(dfs["cm_norm"], dfs["cm_norm_masked"], dpi=72, display=False)
for logger in self.logger:
if isinstance(logger, pl.loggers.wandb.WandbLogger):
import wandb
logger.experiment.log(
{
"Confusion matrix (Val)": wandb.Image(
cm_chart,
caption=f"CM-Val-Norm-{self.trainer.global_step}",
)
},
commit=False,
)
def __init__(self, src='Hand', s=1E-5):
P = np.load(src+'.npz'); self.data = []
for i,(k,v) in enumerate(P.items()):
self.data += [(i,p.astype('float32')) for p in v]
# x_transform: flip + scale, keypoint jitter with magnitude s/2
#self.x_transform = lambda x: torch.ravel(torch.randn(2)*(x+(torch.rand(x.shape)-0.5)*s))
self.x_transform = lambda x: torch.ravel(torch.randn(2)*(1+(torch.rand(x.shape)-0.5)*s)*x)
self.y_transform = None # lambda x: torch.tensor(x)
self.cls = [k for k in P]; print(self.cls)
def generate_coordinates(n: int) -> torch.Tensor:
'''
Genearates grid of 2D coordinates [0, n]x[n, 0]
params
-----------
n: int
Number of 2d points
returns:
-----------
coord_abs: torch.ndarray
image coordinates of (n**2) x 2 size
'''
#meshgrid of torch uses ij so lets just use it
#rather than use np meshgrid and change the
# indexing to ij
with torch.no_grad():
rows, cols = torch.meshgrid(torch.arange(n), torch.arange(n))
# i , j format coordinates
coords_abs = torch.stack(
[torch.ravel(rows), torch.ravel(cols)], axis=-1)
return coords_abs
def ravel(self, tensor):
"""
Return a flattened view of the tensor, not a copy.
Example:
>>> import pyhf
>>> pyhf.set_backend("pytorch")
>>> tensor = pyhf.tensorlib.astensor([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]])
>>> pyhf.tensorlib.ravel(tensor)
tensor([1., 2., 3., 4., 5., 6.])
Args:
tensor (Tensor): Tensor object
Returns:
`torch.Tensor`: A flattened array.
"""
return torch.ravel(tensor)
def other_ops(self):
a = torch.randn(4)
b = torch.randn(4)
c = torch.randint(0, 8, (5, ), dtype=torch.int64)
e = torch.randn(4, 3)
f = torch.randn(4, 4, 4)
size = [0, 1]
dims = [0, 1]
return (
torch.atleast_1d(a),
torch.atleast_2d(a),
torch.atleast_3d(a),
torch.bincount(c),
torch.block_diag(a),
torch.broadcast_tensors(a),
torch.broadcast_to(a, (4)),
# torch.broadcast_shapes(a),
torch.bucketize(a, b),
torch.cartesian_prod(a),
torch.cdist(e, e),
torch.clone(a),
torch.combinations(a),
torch.corrcoef(a),
# torch.cov(a),
torch.cross(e, e),
torch.cummax(a, 0),
torch.cummin(a, 0),
torch.cumprod(a, 0),
torch.cumsum(a, 0),
torch.diag(a),
torch.diag_embed(a),
torch.diagflat(a),
torch.diagonal(e),
torch.diff(a),
torch.einsum("iii", f),
torch.flatten(a),
torch.flip(e, dims),
torch.fliplr(e),
torch.flipud(e),
torch.kron(a, b),
torch.rot90(e),
torch.gcd(c, c),
torch.histc(a),
torch.histogram(a),
torch.meshgrid(a),
torch.lcm(c, c),
torch.logcumsumexp(a, 0),
torch.ravel(a),
torch.renorm(e, 1, 0, 5),
torch.repeat_interleave(c),
torch.roll(a, 1, 0),
torch.searchsorted(a, b),
torch.tensordot(e, e),
torch.trace(e),
torch.tril(e),
torch.tril_indices(3, 3),
torch.triu(e),
torch.triu_indices(3, 3),
torch.vander(a),
torch.view_as_real(torch.randn(4, dtype=torch.cfloat)),
torch.view_as_complex(torch.randn(4, 2)),
torch.resolve_conj(a),
torch.resolve_neg(a),
)
def degrees(M):
signal = torch.reshape(torch.ones(M.shape[0]), (-1, 1))
index, values = torch_sparse.transpose(M.index, M.values, M.shape[0],
M.shape[1])
return torch.ravel(
torch_sparse.spmm(index, values, M.shape[1], M.shape[0], signal))
def conv(signal, M):
signal = torch.reshape(signal, (-1, 1))
return torch.ravel(
torch_sparse.spmm(M.index, M.values, M.shape[0], M.shape[1], signal))
def test_epoch_end(self, outputs: Dict[str, Any]):
# --- original cm
prediction = torch.cat([tmp["prediction"] for tmp in outputs])
target = torch.cat([tmp["target"] for tmp in outputs])
# --- masked cm, flatten all tensors, subset by forest pixels
lu = torch.ravel(torch.cat([tmp["lu"] for tmp in outputs]))
prediction_masked = torch.ravel(prediction)[lu == 1]
target_masked = torch.ravel(target)[lu == 1]
confusion_matrix = torchmetrics.functional.confusion_matrix(
prediction, target, normalize="true", num_classes=len(self.classes)
)
confusion_matrix_px = torchmetrics.functional.confusion_matrix(
prediction, target, num_classes=len(self.classes)
)
confusion_matrix_masked = torchmetrics.functional.confusion_matrix(
prediction_masked,
target_masked,
normalize="true",
num_classes=len(self.classes),
)
confusion_matrix_masked_px = torchmetrics.functional.confusion_matrix(
prediction_masked, target_masked, num_classes=len(self.classes)
)
dfs = {}
for label, cm in zip(
["cm_norm", "cm_px", "cm_norm_masked", "cm_px_masked"],
[
confusion_matrix,
confusion_matrix_px,
confusion_matrix_masked,
confusion_matrix_masked_px,
],
):
dfs[label] = pd.DataFrame(
cm.detach().cpu().numpy(),
index=self.classes,
columns=self.classes,
)
cm_chart = show_cm(dfs["cm_norm"], dfs["cm_norm_masked"], dpi=72, display=False)
cm_chart_px = show_cm(dfs["cm_px"], dfs["cm_px_masked"], dpi=72, display=False)
for logger in self.logger:
if isinstance(logger, pl.loggers.wandb.WandbLogger):
import wandb
logger.experiment.log(
{
"Confusion Matrix (Test) - Normalized": wandb.Image(
cm_chart,
caption=f"CM-Test-Norm-{self.trainer.global_step}",
),
"Confusion Matrix (Test) - Pixel": wandb.Image(
cm_chart_px,
caption=f"CM-Test-Px-{self.trainer.global_step}",
),
},
commit=False,
)
log.info(f"CM - DEFAULT - NORMALIZED: {dfs['cm_norm'].to_string()}")
log.info(f"CM - FORESTONLY - NORMALIZED: {dfs['cm_norm_masked'].to_string()}")
log.info(f"CM - DEFAULT - PIXEL: {dfs['cm_px'].to_string()}")
log.info(f"CM - FORESTONLY - PIXEL: {dfs['cm_px_masked'].to_string()}")
x_img = torch.rand(64)
print(x_img.shape)
x_image_reshaped = x_img.reshape(8, 8)
print(x_image_reshaped.shape)
plt.imshow(x_image_reshaped, cmap='gray')
plt.show()
''' Make 3 channel image '''
channels = 3
image_size = 1728
x_rgb_img = torch.randint(0, 255, (1728, 1))
print(x_rgb_img.shape)
x_rgb_img = torch.ravel(x_rgb_img).reshape(3, 24, 24)
print(x_rgb_img.shape)
x_rgb_img = x_rgb_img.permute(2, 1, 0)
print(x_rgb_img.shape)
# matplot lib cannot display image as channel first
# so we need to reshape the image to channel at end
plt.imshow(x_rgb_img)
plt.show()
# using numpy
x = np.random.random((10, 10, 3))
print(f'numpy array x shape: {x.shape}')
print(x)
plt.imshow(x)
plt.show()
