def reshape(data: dict) -> (np.array, np.array):
length = sum(len(lst) for lst in data.values())
# labels = np.empty(length, dtype='<U1')
# y = np.ndarray(shape =(45, length), dtype='int32')
y = [None] * length
labels = [None] * length
i = 0
for label, pictures in data.items():
for picture in pictures:
labels[i] = label
y[i] = picture.flatten()
i += 1
return np.array(y), np.array(labels)
def process(data):
links = []
titles = []
misc_data = []
for i in data.values():
links.append(i[1])
titles.append(i[0])
misc_data.append(i[2])
df = pd.DataFrame({'name of painting':titles,
'img link':links,
'info':misc_data},
index=[list(data.keys())])
l = []
for i in df['info']:
s_l = []
for j in range(0,len(i),3):
s_l.append(i[j+1:j+3])
l.append(s_l)
df['l'] = l
#Index by each painting, and not each era
list_one = []
for i in range(len(df['name of painting'])): #go through each era
list_two = []
for j in range(len(df['name of painting'][i])): #go through each list
painting = df['name of painting'][i][j]
era = list(df.index[i])[0]
link = df['img link'][i][j]
moreinfo = []
for k in df['l'][i][j]:
moreinfo.append(k)
list_two.append([era,painting,link,moreinfo[0]])
list_one.append(list_two)
tolist = [j for i in list_one for j in i ]
big_df = pd.DataFrame(tolist)
big_df = big_df.rename(columns={0:'era', 1:'painting', 2:'url', 3:'metadata'})
#painting or drawing
paintdraw = []
for i in big_df['metadata']:
if 'paint' in i or 'canvas' in i:
paintdraw.append('Painting')
elif 'paper' in i:
paintdraw.append('Drawing')
else:
paintdraw.append('N/A')
big_df['painting or drawing'] = paintdraw
big_df['metadata'] = big_df['metadata'].str.replace('\n','').str.replace(' ','')
big_df['metadata'] = big_df['metadata'].str.split(',')
#separate type and dimensions
big_df["type"] = big_df["metadata"].str[0]
big_df['dimensions'] = big_df['metadata'].str[1]
big_df = big_df.drop(columns=['metadata'])
return big_df
#embedding.append(emb_array)
#ID.append(i)
#real_ID = ID[i]
real_ID = i
to_json(i, emb_array)
else:
data = from_json()
for k in range(0, len(data)):
dist = np.sqrt(
np.sum(
np.square(
np.subtract(
emb_array,
list(data.values())[k]))))
if (dist < 1.0):
to_json(k, emb_array)
#embedding[k] = emb_array
#real_ID = ID[k]
real_ID = k
present = True
break
if (present == False):
index = int(len(data) + 1)
to_json(index, emb_array)
#embedding.append(emb_array)
#ID.append(int(len(ID) + 1))
#real_ID = int(len(ID) + 1)
real_ID = index