def testPoolingMultiple(self):
height = 10
width = 10
data = np.random.rand(height, width, 3)
data_2d = data.reshape(100,3)
grids = [(2,2), (3,3), (2,3), (4,4), (2,4), (5,5)]
for grid in grids:
index = np.zeros((height, width), dtype=int)
for i in range(height):
for j in range(width):
h_id = i * grid[0] / height;
w_id = j * grid[1] / width;
index[i,j] = h_id * grid[1] + w_id
index = index.flatten()
num_pool = grid[0]*grid[1]
pooled_max = np.hstack(
[data_2d[index==i].max(axis=0) for i in range(num_pool)])
pooled_ave = np.hstack(
[data_2d[index==i].mean(axis=0) for i in range(num_pool)])
pooled_rms = np.hstack(
[(data_2d[index==i]**2).mean(axis=0) \
for i in range(num_pool)])
pooled_rms = np.sqrt(pooled_rms)
np.testing.assert_almost_equal(pooled_max,
cpputil.fastpooling(data, grid, 'max').flatten())
np.testing.assert_almost_equal(pooled_ave,
cpputil.fastpooling(data, grid, 'ave').flatten())
np.testing.assert_almost_equal(pooled_rms,
cpputil.fastpooling(data, grid, 'rms').flatten())
def testPoolingSingle(self):
data = np.random.rand(10,10,3)
data_2d = data.reshape((100,3))
pooled_max = data_2d.max(axis=0)
pooled_ave = data_2d.mean(axis=0)
pooled_rms = np.sqrt((data_2d**2).mean(axis=0))
np.testing.assert_almost_equal(pooled_max,
cpputil.fastpooling(data, (1,1), 'max').flatten())
np.testing.assert_almost_equal(pooled_ave,
cpputil.fastpooling(data, (1,1), 'ave').flatten())
np.testing.assert_almost_equal(pooled_rms,
cpputil.fastpooling(data, (1,1), 'rms').flatten())
def process(self, image, out = None):
if not (image.flags['C_CONTIGUOUS'] and image.dtype == np.float64):
logging.warning("Warning: the image is not contiguous.")
image = np.ascontiguousarray(image, dtype=np.float64)
# do fast pooling
grid = self.specs['grid']
if type(grid) is int:
grid = (grid, grid)
self.specs['grid'] = grid
out = cpputil.fastpooling(image, grid, self.specs['method'], out = out)
return out
def testPoolingShapes(self):
heights = [16, 31, 32, 33, 37, 40]
widths = [16, 31, 32, 33, 37, 40]
channels = [1, 3, 5, 10]
grids = [(2, 2), (3, 3), (2, 3), (4, 4), (2, 4), (5, 5)]
for height in heights:
for width in widths:
for channel in channels:
data = np.random.rand(height, width, channel)
for grid in grids:
shape = cpputil.fastpooling(data, grid, 'max').shape
self.assertEqual(shape, grid + (channel,))