- the plot format follows the style of ann-benchmarks
- X axis is recall@k, Y axis is queries/second
- for the same recall@k, closer to topside edge is better
- for the same queries/second, closer to the right side edge is better
- each node on a curve represents a different
n_probe, starting from 1 at the left most node, and multiplied by 2 at every next node. (n_probe = 1,2,4,8,16,32,64,128)
- number of training vectors: 500,000
- number of base vectors: 1,000,000
- number of query vectors: 1,000
- vector dimentions: 960
- runtime is average of 30 runs
Click here to see details:
Click here to see details:
