AdaSample: Adaptive Sampling of Hard Positives for Descriptor Learning
This work addresses a bottleneck in descriptor learning for computer vision by improving batch construction, though it is incremental as it builds on existing triplet loss methods.
The paper tackles the problem of constructing informative batches for triplet loss in local descriptor learning by proposing AdaSample, an adaptive online batch sampler that samples hard positives based on informativeness, resulting in significant and consistent performance gains on standard benchmarks.
Triplet loss has been widely employed in a wide range of computer vision tasks, including local descriptor learning. The effectiveness of the triplet loss heavily relies on the triplet selection, in which a common practice is to first sample intra-class patches (positives) from the dataset for batch construction and then mine in-batch negatives to form triplets. For high-informativeness triplet collection, researchers mostly focus on mining hard negatives in the second stage, while paying relatively less attention to constructing informative batches. To alleviate this issue, we propose AdaSample, an adaptive online batch sampler, in this paper. Specifically, hard positives are sampled based on their informativeness. In this way, we formulate a hardness-aware positive mining pipeline within a novel maximum loss minimization training protocol. The efficacy of the proposed method is evaluated on several standard benchmarks, where it demonstrates a significant and consistent performance gain on top of the existing strong baselines.