Deep-1M

The 'Deep1M' dataset is a benchmark that contains a large collection of deep learning embeddings used to evaluate approximate $k$-nearest neighbor search methods in vector databases.

Papers using Deep-1M (3)

Approximate Diverse $k$-nearest Neighbor Search in Vector Database2025Qinco2: Vector Compression and Search with Improved Implicit Neural Codebooks2025Residual Quantization with Implicit Neural Codebooks2024