def test_multi_class7(
self): #Test based on small piece of Severstal dataset
TESTDATA = StringIO("""ImageId,ClassId
ff6e35e0a.jpg,1
ff6e35e0a.jpg,2
ff933e271.jpg,3
ff96dfa95.jpg,3
ff9923932.jpg,3
ff9d46e95.jpg,4
ffb48ee43.jpg,3
ffbd081d5.jpg,3
ffc9fdf70.jpg,3
ffcf72ecf.jpg,3
fff02e9c5.jpg,3
fffe98443.jpg,3
ffff4eaa8.jpg,3
ffffd67df.jpg,3
00031f466.jpg,
000418bfc.jpg,
000789191.jpg,
""")
df = pd.read_csv(TESTDATA, sep=",")
ds = image_datasets.MultiClassClassificationDataSet([], df, "ImageId",
"ClassId")
ds.get_value = dummyImage
self.assertEqual(len(ds.classes), 4)
encoded = ds.encode(ds, True)
df['ClassId'] = df['ClassId']
#df['ClassId'] = df['ClassId'].str.strip()
rs = encoded['ClassId'].values == df['ClassId'].values
pass
def test_multi_class3(self):
TESTDATA = StringIO("""Image,Clazz
a1,
a2, X_Y
a3, X
a4, X_Z
""")
df = pd.read_csv(TESTDATA, sep=",")
ds = image_datasets.MultiClassClassificationDataSet([], df, "Image",
"Clazz")
ds.get_value = dummyImage
self.assertEqual(ds[0].y[0], False)
self.assertEqual(ds[2].y[0], True)
encoded = ds.encode(ds, True)
df['Clazz'] = df['Clazz'].str.strip()
rs = encoded["Clazz"].values == df["Clazz"].values
self.assertEqual(rs.sum(), 4, "")
pass
def test_multi_class4(self):
TESTDATA = StringIO("""Image,A,B
a1, 0, 0
a2, 1, 0
a3, 0 ,1
a4, 0 ,0
""")
df = pd.read_csv(TESTDATA, sep=",")
ds = image_datasets.MultiClassClassificationDataSet([], df, "Image",
"A|B")
ds.get_value = dummyImage
self.assertEqual(ds[0].y[0], False)
self.assertEqual(ds[2].y[1], True)
encoded = ds.encode(ds, True)
df['A'] = df['A']
rs = encoded["A"].values == df["A"].values
self.assertEqual(rs.sum(), 4, "")
pass
def get_classify():
return image_datasets.MultiClassClassificationDataSet(
["test_images", "train_images"], "classify.csv", "ImageId", "ClassId")
