Exploring Cross-Domain Few-Shot Classification via Frequency-Aware Prompting

Authors: Tiange Zhang, Qing Cai, Feng Gao, Lin Qi, Junyu Dong

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on CDFSL benchmarks demonstrate the effectiveness of our proposed method as well as robustly improve the performance of existing CD-FLS methods.
Researcher Affiliation Academia Faculty of Computer Science and Technology, Ocean University of China zhangtiange@stu.ouc.edu.cn, {cq, gaofeng, qilin, dongjunyu}@ouc.edu.cn
Pseudocode No Other details can be viewed at the Algorithm. 1 in supplements.
Open Source Code Yes Resources at https://github.com/tinkez/FAP CDFSC.
Open Datasets Yes We conduct extensive experiments under the very strict cross-domain few-shot learning settings where only one single source dataset is used for meta-training, i.e., mini Imagne Net [Vinyals et al., 2016]. There are two CD-FSL benchmarks used in the meta-testing phase for evaluation. One is from the FWT benchmark [Tseng et al., 2020] which includes CUB [Welinder et al., 2010], Cars [Krause et al., 2013], Places [Zhou et al., 2017], and Plantae [Van Horn et al., 2018]. The other one is the BSCD-FSL benchmark [Guo et al., 2020] which consists of Chest X, ISIC, Euro SAT, and Crop Disease.
Dataset Splits Yes For all experiments, we select the model with the best validation accuracy on mini-Image Net for the next testing on eight target datasets.
Hardware Specification Yes All experiments reported are conducted on the RTX 2080Ti GPUs.
Software Dependencies No The paper mentions using 'Adam' as an optimizer and 'Res Net-10' as a feature extractor, but it does not specify any software names with version numbers (e.g., Python, PyTorch, TensorFlow versions, or specific library versions).
Experiment Setup Yes For a fair comparison, Res Net-10 is selected as the feature extractor and the optimizer is Adam with a fixed learning rate α = 0.001. The iteration number for early stopping is kept as Tmax = 5 in all experiments and the gradient ascending step takes a learning rate β from {20, 40, 60, 80}. We maintain the Random Convolution layer applied in [Wang and Deng, 2021] with a probability (1 p), where p is selected from {0.5, 0.6, 0.7}. The filter size k is randomly sampled from the candidate pool K = {1, 3, 5, 7, 11, 15} and the Xavier normal distribution [Glorot and Bengio, 2010] is used to initialize the layer weights. Both stride and padding sizes are determined to keep the image size unchanged.