This paper addresses the problem of large-scale image search, focusing on improving the bag-of-features (BOF) representation. The authors propose two main contributions: Hamming Embedding (HE) and Weak Geometric Consistency (WGC). HE refines visual words by adding binary signatures, enhancing the precision of descriptor matching. WGC integrates weak geometric consistency constraints within the inverted file system, filtering out descriptors that are inconsistent in terms of angle and scale. These methods are integrated efficiently into the inverted file structure, allowing for fast and accurate search across large datasets. Experiments on a dataset of one million images demonstrate significant improvements in retrieval accuracy, with HE and WGC showing complementary benefits. The paper also discusses the complexity of these methods and their performance on various datasets, including the INRIA Holidays dataset and the Oxford5k dataset. The results highlight the effectiveness of the proposed techniques in enhancing the performance of large-scale image search.This paper addresses the problem of large-scale image search, focusing on improving the bag-of-features (BOF) representation. The authors propose two main contributions: Hamming Embedding (HE) and Weak Geometric Consistency (WGC). HE refines visual words by adding binary signatures, enhancing the precision of descriptor matching. WGC integrates weak geometric consistency constraints within the inverted file system, filtering out descriptors that are inconsistent in terms of angle and scale. These methods are integrated efficiently into the inverted file structure, allowing for fast and accurate search across large datasets. Experiments on a dataset of one million images demonstrate significant improvements in retrieval accuracy, with HE and WGC showing complementary benefits. The paper also discusses the complexity of these methods and their performance on various datasets, including the INRIA Holidays dataset and the Oxford5k dataset. The results highlight the effectiveness of the proposed techniques in enhancing the performance of large-scale image search.