def process_eval_image(
cfg,
fname_in,
roi,
fname_out,
spatial_levels,
image_helper,
model,
pca,
eval_dataset_name,
verbose=False,
):
if is_revisited_dataset(eval_dataset_name):
img = image_helper.load_and_prepare_revisited_image(fname_in, roi=roi)
elif is_instre_dataset(eval_dataset_name):
img = image_helper.load_and_prepare_instre_image(fname_in)
else:
img = image_helper.load_and_prepare_image(fname_in, roi=roi)
v = torch.autograd.Variable(img.unsqueeze(0))
vc = v.cuda()
# the model output is a list always.
activation_map = model(vc)[0].cpu()
if verbose:
print(f"Eval image raw activation map shape: { activation_map.shape }")
# process the features: rmac | l2 norm
if cfg.IMG_RETRIEVAL.FEATS_PROCESSING_TYPE == "rmac":
descriptors = get_rmac_descriptors(
activation_map,
spatial_levels,
pca=pca,
normalize=cfg.IMG_RETRIEVAL.NORMALIZE_FEATURES,
)
elif cfg.IMG_RETRIEVAL.FEATS_PROCESSING_TYPE == "gem":
descriptors = gem(
activation_map,
p=cfg.IMG_RETRIEVAL.GEM_POOL_POWER,
add_bias=True,
)
else:
descriptors = activation_map
# Optionally l2 normalize the features.
if (cfg.IMG_RETRIEVAL.NORMALIZE_FEATURES
and cfg.IMG_RETRIEVAL.FEATS_PROCESSING_TYPE != "rmac"):
# RMAC performs normalization within the algorithm, hence we skip it here.
descriptors = l2n(descriptors, dim=1)
# Optionally apply pca.
if pca and cfg.IMG_RETRIEVAL.FEATS_PROCESSING_TYPE != "rmac":
# RMAC performs pca within the algorithm, hence we skip it here.
descriptors = pca.apply(descriptors)
if fname_out:
save_file(descriptors.data.numpy(), fname_out, verbose=False)
return descriptors.data.numpy()
def process_train_image(i, out_dir, verbose=False):
if i % LOG_FREQUENCY == 0:
logging.info(f"Train Image: {i}"),
fname_out = None
if out_dir:
fname_out = f"{out_dir}/{i}.npy"
if fname_out and PathManager.exists(fname_out):
feat = load_file(fname_out)
train_features.append(feat)
else:
fname_in = train_dataset.get_filename(i)
if is_revisited_dataset(train_dataset_name):
img = image_helper.load_and_prepare_revisited_image(fname_in,
roi=None)
elif is_whiten_dataset(train_dataset_name):
img = image_helper.load_and_prepare_whitening_image(fname_in)
else:
img = image_helper.load_and_prepare_image(fname_in, roi=None)
v = torch.autograd.Variable(img.unsqueeze(0))
vc = v.cuda()
# the model output is a list always.
activation_map = model(vc)[0].cpu()
if verbose:
print(
f"Train Image raw activation map shape: { activation_map.shape }"
)
# once we have the features,
# we can perform: rmac | gem pooling | l2 norm
if cfg.IMG_RETRIEVAL.FEATS_PROCESSING_TYPE == "rmac":
descriptors = get_rmac_descriptors(
activation_map,
spatial_levels,
normalize=cfg.IMG_RETRIEVAL.NORMALIZE_FEATURES,
)
elif cfg.IMG_RETRIEVAL.FEATS_PROCESSING_TYPE == "gem":
descriptors = gem(
activation_map,
p=cfg.IMG_RETRIEVAL.GEM_POOL_POWER,
add_bias=True,
)
else:
descriptors = activation_map
# Optionally l2 normalize the features.
if (cfg.IMG_RETRIEVAL.NORMALIZE_FEATURES
and cfg.IMG_RETRIEVAL.FEATS_PROCESSING_TYPE != "rmac"):
# RMAC performs normalization within the algorithm, hence we skip it here.
descriptors = l2n(descriptors, dim=1)
if fname_out:
save_file(descriptors.data.numpy(), fname_out, verbose=False)
train_features.append(descriptors.data.numpy())
def gem_pool_and_save_features(features, p, add_bias, gem_out_fname):
if PathManager.exists(gem_out_fname):
logging.info("Loading train GeM features...")
features = load_file(gem_out_fname)
else:
logging.info(f"GeM pooling features: {features.shape}")
features = l2n(gem(features, p=p, add_bias=True))
save_file(features, gem_out_fname)
logging.info(f"Saved GeM features to: {gem_out_fname}")
return features