Weak-shot Fine-grained Classification via Similarity Transfer

Authors: Junjie Chen, Li Niu, Liu Liu, Liqing Zhang

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Comprehensive experiments demonstrate the effectiveness of our weak-shot setting and our Sim Trans method. Datasets and codes are available at https://github.com/bcmi/Sim Trans-Weak-Shot-Classification.
Researcher Affiliation Academia Junjie Chen, Li Niu , Liu Liu, Liqing Zhang Mo E Key Lab of Artificial Intelligence, Department of Computer Science and Engineering, Shanghai Jiao Tong University {chen.bys, ustcnewly, shirlley}@sjtu.edu.cn, zhang-lq@cs.sjtu.edu.cn
Pseudocode No The paper describes its methods in prose and with diagrams but does not contain structured pseudocode or algorithm blocks (clearly labeled algorithm sections or code-like formatted procedures).
Open Source Code Yes Datasets and codes are available at https://github.com/bcmi/Sim Trans-Weak-Shot-Classification.
Open Datasets Yes We conduct experiments based on three fine-grained datasets: Comp Cars [52] (Car for short), CUB [48], and FGVC [26].
Dataset Splits No Table 1 provides 'Train' and 'Test' statistics for the datasets, and the text mentions 'base training/test set' and 'novel training set'. However, there is no explicit separate 'validation' split with specific counts or percentages provided for the datasets, other than mentioning 'cross-validation' for hyperparameter tuning.
Hardware Specification No The paper does not provide any specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments.
Software Dependencies No The paper mentions using ResNet50 as a backbone but does not specify any software dependencies with version numbers (e.g., Python, PyTorch/TensorFlow versions, CUDA versions).
Experiment Setup Yes The classification loss and the adversarial loss are balanced with a hyper-parameter β, set as 0.1 via cross-validation. ... where α is a hyper-parameter set as 0.1 by cross-validation. ... We use Cm = 10 and M = 100 for both training and testing of Sim Net.