Super-Class Guided Transformer for Zero-Shot Attribute Classification
Authors: Sehyung Kim, Chanhyeong Yang, Jihwan Park, Taehoon Song, Hyunwoo J. Kim
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate that Suga Former achieves state-of-the-art performance across three widely-used attribute classification benchmarks under zero-shot, and cross-dataset transfer settings. |
| Researcher Affiliation | Academia | Department of Computer Science and Engineering, Korea University EMAIL |
| Pseudocode | No | The paper describes methods using equations and diagrams (Figure 2 and Figure 3) but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper states: "Further implementation details, including hyperparameters, can be found in the supplementary materials." This does not explicitly mention the release of source code for the methodology described in this paper, nor does it provide a link to a code repository. |
| Open Datasets | Yes | The effectiveness of our Suga Former framework is validated using three attribute classification datasets: VAW (Pham et al. 2021), LSA (Pham et al. 2022), and OVAD (Bravo et al. 2023). |
| Dataset Splits | Yes | For zero-shot attribute classification, following prior work (Chen et al. 2023a), we use half of the tail attributes and 15% of medium attributes as novel classes, resulting in 79 novel classes and 541 base classes. ... We evaluate under zero-shot setting (common-to-rare), with 5526 base attributes and 4012 novel attributes. ... The dataset organizes attributes into three subsets: head (15 frequently occurring attributes), medium (53 moderately frequent attributes), and tail (49 rare attributes) based on thresholds defined by attribute annotation frequency. |
| Hardware Specification | Yes | All training and evaluations are conducted on NVIDIA RTX 3090. |
| Software Dependencies | No | The paper mentions models and backbones like "Vi T-g/14" and "Q-Former", but does not provide specific version numbers for software dependencies such as programming languages, libraries, or frameworks (e.g., Python, PyTorch, CUDA). |
| Experiment Setup | No | Further implementation details, including hyperparameters, can be found in the supplementary materials. |