def main():
cfg = configdataset(test_dataset, data_root)
ranks = global_search(GLOBAL_FEATURE_PATH)
reportMAP(test_dataset, cfg, ranks)
#_, ranks_after_gv = rerankGV(cfg, LOCAL_FEATURE_PATH, ranks)
_, ranks_after_gv = rerankGV_mulprocess(cfg, LOCAL_FEATURE_PATH, ranks)
#np.save("ranks_after_gv.npy", ranks_after_gv)
reportMAP(test_dataset, cfg, ranks_after_gv)
print("Done!")
def rankASMK():
cfg = configdataset(test_dataset, data_root)
ranks = global_search(GLOBAL_FEATURE_PATH)
reportMAP(test_dataset, cfg, ranks)
with open(ASMK_SCORE_PATH, "rb") as fin:
scores = pickle.load(fin)
print("scores", scores.shape)
ranks = ranks.T
ranks_after = ranks
for i in range(ranks.shape[0]):
asmk_scores = scores[i, ranks[i, :NUM_RERANK]]
ranks_after[i, :NUM_RERANK] = ranks[i, np.argsort(-1 * asmk_scores)]
reportMAP(test_dataset, cfg, ranks_after.T)
print("Done!")
#---------------------------------------------------------------------
# Read images
#---------------------------------------------------------------------
def pil_loader(path):
# open path as file to avoid ResourceWarning (https://github.com/python-pillow/Pillow/issues/835)
with open(path, 'rb') as f:
img = Image.open(f)
return img.convert('RGB')
print('>> {}: Processing test dataset...'.format(test_dataset))
# config file for the dataset
# separates query image list from database image list, if revisited protocol used
cfg = configdataset(test_dataset, os.path.join(data_root, 'datasets'))
# query images
for i in np.arange(cfg['nq']):
qim = pil_loader(cfg['qim_fname'](cfg, i)).crop(cfg['gnd'][i]['bbx'])
##------------------------------------------------------
## Perform image processing here, eg, feature extraction
##------------------------------------------------------
print('>> {}: Processing query image {}'.format(test_dataset, i + 1))
for i in np.arange(cfg['n']):
im = pil_loader(cfg['im_fname'](cfg, i))
##------------------------------------------------------
## Perform image processing here, eg, feature extraction
##------------------------------------------------------
print('>> {}: Processing database image {}'.format(test_dataset, i + 1))
if i==0:
code=gd
print(111)
else:
code=np.concatenate((code,gd),axis=0)
code =np.array(code)
code -= np.mean(code, axis=0)
code /= np.std(code, axis=0)
print(code.shape)
sim = np.dot(code, query.T)
ranks = np.argsort(-sim, axis=0)
dataset='roxford5k'
INPUT_PATH = '/home/yangyc/revisitop-master/data/datasets/'
cfg = configdataset(dataset,INPUT_PATH)
gnd = cfg['gnd']
# evaluate ranks
ks = [1, 5, 10]
# search for easy
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['easy']])
g['junk'] = np.concatenate([gnd[i]['junk'], gnd[i]['hard']])
gnd_t.append(g)
mapE, apsE, mprE, prsE = compute_map(ranks, gnd_t, ks)
# Check, and, if necessary, download distractor dataset
download_distractors(data_root)
# Set up the dataset name
distractors_dataset = 'revisitop1m'
# ---------------------------------------------------------------------
# Read images
# ---------------------------------------------------------------------
def pil_loader(path):
# to avoid crashing for truncated (corrupted images)
ImageFile.LOAD_TRUNCATED_IMAGES = True
# open path as file to avoid ResourceWarning
# (https://github.com/python-pillow/Pillow/issues/835)
with open(path, 'rb') as f:
img = Image.open(f)
return img.convert('RGB')
print('>> {}: Processing dataset...'.format(distractors_dataset))
# config file for the dataset
cfg = configdataset(distractors_dataset, os.path.join(data_root, 'datasets'))
for i in np.arange(cfg['n']):
im = pil_loader(cfg['im_fname'](cfg, i))
##------------------------------------------------------
## Perform image processing here, eg, feature extraction
##------------------------------------------------------
print('>> {}: Processing image {}'.format(distractors_dataset, i + 1))