add_with_ids not implemented for this type of index

[foocker]

Problem:

RuntimeError: Error in virtual void faiss::IndexBinary::add_with_ids(faiss::IndexBinary::idx_t, const uint8_t*, const idx_t*) at IndexBinary.cpp:35: add_with_ids not implemented for this type of index

Running on:

• CPU using yes.
• GPU version yes, but not use.

Interface:

• C++
• Python yes

Reproduction instructions

simple code as follows:
index = faiss.IndexBinaryFlat(dimensions*8) #dimensions=32
features = []
ids = []
ids_count =0
for loop:
ret, feature = way_feature(way, file_name) # eg, cv2.ORB_create()...
ids_list = np.linspace(ids_count, ids_count, num=feature.shape[0], dtype="int64")
features.append(feature)
ids.append(ids_list)

features = np.vstack(features)
ids = np.hstack(ids)
if features.any():
if not index.is_trained:
index.train(features)
index.add_with_ids(features, ids) # change

[mdouze]

add_with_ids is not supported for Flat indexes.
See https://github.com/facebookresearch/faiss/wiki/Guidelines-to-choose-an-index#then-flat

[foocker]

add_with_ids is not supported for Flat indexes.
See https://github.com/facebookresearch/faiss/wiki/Guidelines-to-choose-an-index#then-flat

thx for your replay.
some points:
surf's feature is float, so i can use "IDMap,Flat", it's work. but orb's feature vector is uint8, i should use binary index, and hamming distance, not cosine or L2, all sence will be add_with_ids.
so i have two questions:
1.binary index, with add_with_ids, using hamming, what index should i choose?
2.when i change orb's feature from surf'feature, surf work well, i got :
TypeError: in method 'Index_add_with_ids', argument 3 of type 'float const *'

[foocker]

add_with_ids is not supported for Flat indexes.
See https://github.com/facebookresearch/faiss/wiki/Guidelines-to-choose-an-index#then-flat

thx for your replay.
some points:
surf's feature is float, so i can use "IDMap,Flat", it's work. but orb's feature vector is uint8, i should use binary index, and hamming distance, not cosine or L2, all sence will be add_with_ids.
so i have two questions:
1.binary index, with add_with_ids, using hamming, what index should i choose?
2.when i change orb's feature from surf'feature, surf work well, i got :
TypeError: in method 'Index_add_with_ids', argument 3 of type 'float const *'

i have test some other binary index, in binary wiki, add_with_ids was not implemented。
but IndexBinaryHash is ok, but , what's means of IndexBinaryHash 's second parameter?
is hash radius?nflip , radius=nflip? ---> i guess is num of b(桶).
but when use IndexBinaryHash write_index, i get:
TypeError: Wrong number or type of arguments for overloaded function 'write_index'.
Possible C/C++ prototypes are:
faiss::write_index(faiss::Index const *,char const *)
faiss::write_index(faiss::Index const *,FILE *)
faiss::write_index(faiss::Index const *,faiss::IOWriter *)

sorry, write_index_binary is ok..

[goldwater668]

@foocker Open CV feature vectors such as surf and sift have been extracted. How to use faiss for vector index? Do you have any reference cases? Thank you!