3rd Place Solution to Google Landmark Recognition Competition 2021
This is an incremental solution for participants in image recognition competitions, specifically targeting landmark identification.
The paper tackled the Google Landmark Recognition 2021 Competition by using an ensemble of embeddings from CNN-, Transformer-, and hybrid-based architectures optimized with ArcFace loss, along with a re-ranking pipeline, achieving a score of 0.489 and third place on the private leaderboard.
In this paper, we show our solution to the Google Landmark Recognition 2021 Competition. Firstly, embeddings of images are extracted via various architectures (i.e. CNN-, Transformer- and hybrid-based), which are optimized by ArcFace loss. Then we apply an efficient pipeline to re-rank predictions by adjusting the retrieval score with classification logits and non-landmark distractors. Finally, the ensembled model scores 0.489 on the private leaderboard, achieving the 3rd place in the 2021 edition of the Google Landmark Recognition Competition.