def evaluate_depth(self, batch):
"""Evaluate batch to produce depth metrics."""
# Get predicted depth
inv_depths = self.model(batch)['inv_depths']
depth = inv2depth(inv_depths[0])
# Post-process predicted depth
batch['rgb'] = flip_lr(batch['rgb'])
inv_depths_flipped = self.model(batch)['inv_depths']
inv_depth_pp = post_process_inv_depth(inv_depths[0],
inv_depths_flipped[0],
method='mean')
depth_pp = inv2depth(inv_depth_pp)
batch['rgb'] = flip_lr(batch['rgb'])
# Calculate predicted metrics
metrics = OrderedDict()
if 'depth' in batch:
for mode in self.metrics_modes:
if self.config.model.loss.mask_ego:
metrics[self.metrics_name +
mode] = compute_ego_depth_metrics(
self.config.model.params,
gt=batch['depth'],
pred=depth_pp if 'pp' in mode else depth,
path_to_ego_masks=batch['path_to_ego_mask'],
use_gt_scale='gt' in mode)
else:
metrics[self.metrics_name + mode] = compute_depth_metrics(
self.config.model.params,
gt=batch['depth'],
pred=depth_pp if 'pp' in mode else depth,
use_gt_scale='gt' in mode)
# Return metrics and extra information
return {'metrics': metrics, 'inv_depth': inv_depth_pp}
def evaluate_depth(self, batch):
"""Evaluate batch to produce depth metrics."""
# Get predicted depth
inv_depths = self.model(batch)["inv_depths"]
save_image(inv_depths[0], "inv_depths.png") # new
depth = inv2depth(inv_depths[0])
# Post-process predicted depth
batch["rgb"] = flip_lr(batch["rgb"])
inv_depths_flipped = self.model(batch)["inv_depths"]
inv_depth_pp = post_process_inv_depth(inv_depths[0],
inv_depths_flipped[0],
method="mean")
depth_pp = inv2depth(inv_depth_pp)
batch["rgb"] = flip_lr(batch["rgb"])
# Calculate predicted metrics
metrics = OrderedDict()
if "depth" in batch:
for mode in self.metrics_modes:
metrics[self.metrics_name + mode] = compute_depth_metrics(
self.config.model.params,
gt=batch["depth"],
pred=depth_pp if "pp" in mode else depth,
use_gt_scale="gt" in mode,
)
# Return metrics and extra information
return {"metrics": metrics, "inv_depth": inv_depth_pp}
def main(args):
# Get and sort ground-truth and predicted files
exts = ('npz', 'png')
gt_files, pred_files = [], []
for ext in exts:
gt_files.extend(glob(os.path.join(args.gt_folder, '*.{}'.format(ext))))
pred_files.extend(
glob(os.path.join(args.pred_folder, '*.{}'.format(ext))))
# Sort ground-truth and prediction
gt_files.sort()
pred_files.sort()
# Loop over all files
metrics = []
progress_bar = tqdm(zip(gt_files, pred_files), total=len(gt_files))
for gt, pred in progress_bar:
# Get and prepare ground-truth and predictions
gt = torch.tensor(load_depth(gt)).unsqueeze(0).unsqueeze(0)
pred = torch.tensor(load_depth(pred)).unsqueeze(0).unsqueeze(0)
# Calculate metrics
metrics.append(
compute_depth_metrics(args,
gt,
pred,
use_gt_scale=args.use_gt_scale))
# Get and print average value
metrics = (sum(metrics) / len(metrics)).detach().cpu().numpy()
names = ['abs_rel', 'sqr_rel', 'rmse', 'rmse_log', 'a1', 'a2', 'a3']
for name, metric in zip(names, metrics):
print('{} = {}'.format(name, metric))
def evaluate_depth_maps(pred_folder, gt_folder, use_gt_scale, **kwargs):
"""
Calculates depth metrics from a folder of predicted and ground-truth depth files
Parameters
----------
pred_folder : str
Folder containing predicted depth maps (.npz with key 'depth')
gt_folder : str
Folder containing ground-truth depth maps (.npz with key 'depth')
use_gt_scale : bool
Using ground-truth median scaling or not
kwargs : dict
Extra parameters for depth evaluation
"""
# Get and sort ground-truth files
gt_files = glob(os.path.join(gt_folder, '*.npz'))
gt_files.sort()
# Get and sort predicted files
pred_files = glob(os.path.join(pred_folder, '*.npz'))
pred_files.sort()
# Prepare configuration
config = Namespace(**kwargs)
# Loop over all files
metrics = []
for gt, pred in zip(gt_files, pred_files):
# Get and prepare ground-truth
gt = np.load(gt)['depth']
gt = torch.tensor(gt).unsqueeze(0).unsqueeze(0)
# Get and prepare predictions
pred = np.load(pred)['depth']
pred = torch.tensor(pred).unsqueeze(0).unsqueeze(0)
# Calculate metrics
metrics.append(
compute_depth_metrics(config, gt, pred, use_gt_scale=use_gt_scale))
# Get and print average value
metrics = (sum(metrics) / len(metrics)).detach().cpu().numpy()
names = ['abs_rel', 'sqr_rel', 'rmse', 'rmse_log', 'a1', 'a2', 'a3']
for name, metric in zip(names, metrics):
print('{} = {}'.format(name, metric))