def test_cpp_extension(self):
self.assertTrue((hashranking.hamming_rank([[1.0]],
[[1.0]]) == [[0]]).all())
self.assertTrue((hashranking.hamming_rank(
[[1.0, -1.0, -1.0], [-1.0, 1.0, -1.0], [1.0, 1.0, -1.0]],
[[1.0, -1.0, -1.0], [1.0, 1.0, -1.0], [-1.0, 1.0, -1.0]]) == [[
0, 1, 2
], [2, 1, 0], [1, 0, 2]]).all())
def test_on_random_64(self):
b1 = np.random.rand(200, 64).astype(np.float32) - 0.5
b2 = np.random.rand(500, 64).astype(np.float32) - 0.5
d1 = hashranking.hamming_rank(b1, b2)
d2 = hashranking.numpy_implementation.hamming_rank(b1, b2)
self.assertTrue((d1 == d2).all())
def test_map_from_rank(self):
db_size = 1000
query_size = 200
class_count = 10
hash_size = 24
top_n = 50
db = np.random.randint(class_count, size=db_size, dtype=np.uint32)
query = np.random.randint(class_count,
size=query_size,
dtype=np.uint32)
hashes_class = np.random.rand(class_count, hash_size).astype(
np.float32) - 0.5
hashes_db = hashes_class[db] + 0.3 * np.random.randn(
db_size, hash_size).astype(np.float32)
hashes_query = hashes_class[query] + 0.3 * np.random.randn(
query_size, hash_size).astype(np.float32)
s = hashranking.numpy_implementation._compute_similarity(db, query)
rank = hashranking.numpy_implementation.hamming_rank(
hashes_query, hashes_db)
mAP_py, p, r = hashranking.numpy_implementation.compute_map_from_rank(
rank, s, top_n)
rank = hashranking.hamming_rank(hashes_query, hashes_db)
mAP_cpp, p, r = hashranking.compute_map_from_rank(rank, s, top_n)
self.assertEqual(mAP_py, mAP_cpp)
