Bag of Tricks and a Strong Baseline for FGVC
Fine-grained visual classification (FGVC), as a subclass classification task under the superclass, brings more challenges. However, in addition to fine-grained features, the FungiCLEF 2022 dataset is also characterized by imbalance and rich meta-information. This motivates us to explore the impact of different methods and components in fine-grained classification on FungiCLEF 2022. We explore the impact of different data augmentations, backbones, loss functions, and attention mechanisms on classification performance. Additionally, we explore different metadata usage scenarios. In the end, we win second place in the CVPR2022 FGVC Workshop FungiCLEF 2022 challenge. Our code is available at https: //github.com/wujiekd/Bag-of-Tricks-and-a-Strong-Baseline-for-Fungi-Fine-Grained-Classification.
PDF