def test_detect_crop(self):
d1 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d1.fit([['posd1'], ['negd1']])
il1 = d1.detect_crop(img1,
classes='1',
threshold=0.6,
post_processing='None',
overlap_threshold=0.0)
assert len(il1) == 2
assert il1[0].shape == (20, 10)
assert il1[1].shape == (20, 10)
assert il1[0].sum() == 200
assert il1[1].sum() == 200
d2 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d2.fit([['posd1'], ['negd1']])
il2 = d2.detect_crop(img2,
classes='1',
threshold=0.6,
post_processing='None',
overlap_threshold=0.0)
assert len(il2) == 2
assert il2[0].shape == (20, 10, 2)
assert il2[1].shape == (20, 10, 2)
assert il2[0].sum() == 400
assert il2[1].sum() == 400
def test_fit(self):
d1 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=0.5,
scale_step=0.5,
max_scale_steps=1)
path1 = [[data['pos1'], data['pos2']],
[data['neg1'], data['neg2'], data['neg3']]]
d1.fit(path1, store_data=True)
assert d1.X.shape == (5, 128)
assert d1.y.shape == (5, )
assert d1.classes_ == ['0', '1']
d2 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=0.5,
scale_step=0.5,
max_scale_steps=1)
path2 = [['pos1', 'pos2'], ['neg1', 'neg2', 'neg3']]
d2.fit(path2, store_data=True)
assert d2.X.shape == (5, 128)
assert d2.y.shape == (5, )
assert d2.classes_ == ['0', '1']
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=0.5,
scale_step=0.5,
max_scale_steps=1)
path3 = {'pos': ['pos1', 'pos2'], 'neg': ['neg1', 'neg2', 'neg3']}
d3.fit(path3, store_data=True)
assert d3.X.shape == (5, 128)
assert d3.y.shape == (5, )
assert d3.classes_ == ['pos', 'neg'] or d3.classes_ == ['neg', 'pos']
d4 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=0.5,
scale_step=0.5,
max_scale_steps=1)
with pytest.raises(TypeError):
d4.fit(1)
d4 = gpod.Detector2D(lo1,
de1,
CL1(), (17, 4),
frame_step=0.5,
scale_step=0.5,
max_scale_steps=1)
path4 = {'pos': ['pos1', 'pos2'], 'neg': ['neg1', 'neg2', 'neg3']}
with pytest.raises(ValueError):
d4.fit(path4)
def test_detect_mark(self):
d1 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d1.fit([['posd1'], ['negd1']])
im1 = d1.detect_mark(img1,
classes='1',
threshold=0.6,
post_processing='None',
overlap_threshold=0.0)
assert isinstance(im1, numpy.ndarray)
assert im1.shape == (100, 100)
assert im1.sum() == 200 + 200 - 56 - 56
d2 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d2.fit([['posd1'], ['negd1']])
im2 = d2.detect_mark(img1,
classes='1',
threshold=0.6,
post_processing='None',
overlap_threshold=0.0,
color=3)
assert isinstance(im2, numpy.ndarray)
assert im2.shape == (100, 100)
assert im2.sum() == 200 + 200 + 56 * 2 + 56 * 2
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d3.fit([['posd1'], ['negd1']])
im3 = d3.detect_mark(img2,
classes='1',
threshold=0.6,
post_processing='None',
overlap_threshold=0.0,
color=(3, 4))
assert isinstance(im3, numpy.ndarray)
assert im3.shape == (100, 100, 2)
assert im3.sum() == 400 + 400 + 56 * 5 + 56 * 5
def test_batch_detect_crop(self):
d1 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d1.fit([['posd1'], ['negd1']])
bimgs = [img1, img1, img1]
lmo1 = d1.batch_detect_crop(bimgs,
classes=None,
threshold=0.0,
post_processing='None',
overlap_threshold=0.0,
flatten=False)
assert isinstance(lmo1, list)
assert len(lmo1) == 3
assert isinstance(lmo1[0], list)
assert isinstance(lmo1[0][0], numpy.ndarray)
d2 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d2.fit([['posd1'], ['negd1']])
bimgs = [img1, img1, img1]
lmo2 = d2.batch_detect_crop(bimgs,
classes=None,
threshold=0.6,
post_processing='None',
overlap_threshold=0.0,
flatten=True)
assert isinstance(lmo2, list)
assert len(lmo2) == 6
assert isinstance(lmo2[0], numpy.ndarray)
assert lmo2[0].shape == (20, 10)
def test_batch_detect(self):
d1 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d1.fit([['posd1'], ['negd1']])
bimgs = [img1, img1, img1]
lmo1 = d1.batch_detect(bimgs,
classes=None,
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
assert isinstance(lmo1, list)
assert len(lmo1) == 3
assert isinstance(lmo1[0], gpod.Mark2D)
def test_create(self):
d1 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=0.5,
scale_step=0.5,
max_scale_steps=1)
d1._check_ldc()
d1._check_format_params()
assert d1._frame == (16, 8)
assert d1._frame_step == (8, 4)
assert d1._max_scale_steps == 1
d2 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=4)
d2._check_ldc()
d2._check_format_params()
assert d2._frame == (16, 8)
assert d2._frame_step == (8, 4)
assert d2._max_scale_steps == 4
with pytest.raises(TypeError):
d3 = gpod.Detector2D(lo1,
CL1(),
CL1(), (16, 8),
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(TypeError):
d3 = gpod.Detector2D(lo1,
de1,
de1, (16, 8),
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(TypeError):
d3 = gpod.Detector2D(CL1(),
de1,
CL1(), (16, 8),
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(TypeError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, ),
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(TypeError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8, 2),
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(TypeError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=(8, ),
scale_step=0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(TypeError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=-1,
scale_step=0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(TypeError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(),
-1,
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(ValueError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=(8, 4),
scale_step=1.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(ValueError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=(8, 4),
scale_step=-0.5,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(ValueError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=(8, 4),
scale_step=0.0,
max_scale_steps=4)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(ValueError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=0)
d3._check_ldc()
d3._check_format_params()
with pytest.raises(ValueError):
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (16, 8),
frame_step=(8, 4),
scale_step=0.5,
max_scale_steps=-6)
d3._check_ldc()
d3._check_format_params()
def test_detect(self):
d1 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d1.fit([['posd1'], ['negd1']])
mo1 = d1.detect(img1,
classes=None,
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
assert isinstance(mo1, gpod.Mark2D)
assert len(mo1) == 180
assert len(mo1['0']) == 90
assert len(mo1[0.4:]) == 6
assert len(mo1[0.6:]) == 2
assert len(mo1[0.4:].apply_nms(0.3)) == 2
assert len(mo1[0.4:]['1'].apply_nms(0.3)) == 2
assert len(mo1['1'][0.4:].apply_nms(0.3)) == 2
lof = mo1['1'][0.4:].apply_nms(0.3).get_list_of_frames()
assert lof[0] == (10, 10, 20, 10, 1.0, '1')
assert lof[1] == (20, 40, 20, 10, 1.0, '1')
d2 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=2)
d2.fit([['posd1'], ['negd1']])
mo2 = d2.detect(img1,
classes=None,
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
assert isinstance(mo2, gpod.Mark2D)
assert len(mo2) == 220
assert len(mo2['0']) == 110
assert len(mo2[0.10:]) == 6 + 4
assert len(mo2[0.13:]) == 6 + 3
assert len(mo2[0.26:]) == 6 + 0
assert len(mo2[0.4:].apply_nms(0.3)) == 2
assert len(mo2[0.4:]['1'].apply_nms(0.3)) == 2
assert len(mo2['1'][0.4:].apply_nms(0.3)) == 2
lof = mo2['1'][0.4:].apply_nms(0.3).get_list_of_frames()
assert lof[0] == (10, 10, 20, 10, 1.0, '1')
assert lof[1] == (20, 40, 20, 10, 1.0, '1')
d3 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=3)
d3.fit([['posd1'], ['negd1']])
mo3 = d3.detect(img1,
classes=None,
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
assert len(mo3) == 224
assert len(mo3['1']) == 112
assert len(mo3[0.4:].apply_nms(0.3)) == 2
assert len(mo3[0.4:]['1'].apply_nms(0.3)) == 2
assert len(mo3['1'][0.4:].apply_nms(0.3)) == 2
lof = mo3['1'][0.4:].apply_nms(0.3).get_list_of_frames()
assert lof[0] == (10, 10, 20, 10, 1.0, '1')
assert lof[1] == (20, 40, 20, 10, 1.0, '1')
d4 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=100)
d4.fit([['posd1'], ['negd1']])
mo4 = d4.detect(img1,
classes=None,
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
assert len(mo4) == 224
assert len(mo4['1']) == 112
assert len(mo4[0.4:].apply_nms(0.3)) == 2
assert len(mo4[0.4:]['1'].apply_nms(0.3)) == 2
assert len(mo4['1'][0.4:].apply_nms(0.3)) == 2
lof = mo4['1'][0.4:].apply_nms(0.3).get_list_of_frames()
assert lof[0] == (10, 10, 20, 10, 1.0, '1')
assert lof[1] == (20, 40, 20, 10, 1.0, '1')
d5 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d5.fit([['posd1'], ['negd1']])
mo5 = d5.detect(img1,
classes=None,
threshold=0.01,
post_processing='NMS',
overlap_threshold=0.3)
assert len(mo5) == 2
lof = mo5.get_list_of_frames()
assert lof[0] == (10, 10, 20, 10, 1.0, '1')
assert lof[1] == (20, 40, 20, 10, 1.0, '1')
d6 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=1)
d6.fit([['posd1'], ['negd1']])
mo6 = d6.detect(img1,
classes='1',
threshold=0.6,
post_processing='None',
overlap_threshold=0.0)
assert len(mo6) == 2
lof = mo6.get_list_of_frames()
assert lof[0] == (10, 10, 20, 10, 1.0, '1')
assert lof[1] == (20, 40, 20, 10, 1.0, '1')
d7 = gpod.Detector2D(lo1,
de1,
CL1(), (20, 10),
frame_step=(10, 10),
scale_step=0.5,
max_scale_steps=2)
d7.fit([['posd1'], ['negd1']])
mo7 = d7.detect(img2,
classes=None,
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
assert isinstance(mo7, gpod.Mark2D)
assert len(mo7) == 220
assert len(mo7['0']) == 110
assert len(mo7[0.10:]) == 6 + 4
assert len(mo7[0.13:]) == 6 + 3
assert len(mo7[0.26:]) == 6 + 0
assert len(mo7[0.4:].apply_nms(0.3)) == 2
assert len(mo7[0.4:]['1'].apply_nms(0.3)) == 2
assert len(mo7['1'][0.4:].apply_nms(0.3)) == 2
lof = mo7['1'][0.4:].apply_nms(0.3).get_list_of_frames()
assert lof[0] == (10, 10, 20, 10, 1.0, '1')
assert lof[1] == (20, 40, 20, 10, 1.0, '1')
mo7 = d7.detect(img2,
classes='1',
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
mo7 = d7.detect(img2,
classes=['1'],
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
mo7 = d7.detect(img2,
classes=set(['1']),
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
with pytest.raises(ValueError):
mo7 = d7.detect(img2,
classes='pos',
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
with pytest.raises(ValueError):
mo7 = d7.detect(img2,
classes=['pos'],
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
with pytest.raises(ValueError):
mo7 = d7.detect(img2,
classes=set(['pos']),
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)
with pytest.raises(ValueError):
mo7 = d7.detect(img2,
classes=1,
threshold=0.0,
post_processing='None',
overlap_threshold=0.0)