HOIMamba: Efficient Mamba-based Disentangled Progressive Learning for HOI Detection
Authors: Yongchao Xu, Jiawei Liu, Sen Tao, Qiang Zhang, Zheng-Jun Zha
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experimental results on two standard benchmarks demonstrate the effectiveness of our HOIMamba. ... Experiments Experimental Settings Datasets and Evaluation Metrics. We evaluate the proposed model on two public benchmarks, HICO-DET (Chao et al. 2018) and V-COCO (Gupta and Malik 2015), and use the Mean Average Precision (m AP) metric on both datasets. ... Comparison with State-of-the-Art Methods ... Ablation study on overall model framework design. |
| Researcher Affiliation | Academia | 1University of Science and Technology of China, Hefei, China 2 University of Chinese Academy of Sciences, Beijing, China EMAIL, EMAIL, EMAIL |
| Pseudocode | No | The paper describes methods using mathematical equations and block diagrams (Figure 3), but does not contain explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain any explicit statement about releasing source code or provide a link to a code repository. |
| Open Datasets | Yes | We evaluate the proposed model on two public benchmarks, HICO-DET (Chao et al. 2018) and V-COCO (Gupta and Malik 2015), and use the Mean Average Precision (m AP) metric on both datasets. |
| Dataset Splits | No | The paper mentions evaluating on HICO-DET and V-COCO and states, 'A detailed description of the datasets and evaluation metrics can be found in the Supplementary Materials.' However, the main text does not explicitly provide specific training/test/validation dataset splits, percentages, or sample counts. |
| Hardware Specification | Yes | All experiments are carried out on 4 RTX3090 GPUs with batch size set to 16. ... All models are tested using one RTX 3090 with an input of 640 640 resolution. |
| Software Dependencies | No | The paper mentions using Adam W for optimization but does not provide specific version numbers for software libraries or frameworks such as PyTorch, TensorFlow, or CUDA. |
| Experiment Setup | Yes | For our progressive learning, the hyperparameters for CESM φ1 and φ2 are set to 0.50 and 0.75, respectively. ... During training, following MUREN (Kim, Jung, and Cho 2023), we set the number of queries Nq to 64, the number of channels Ce and Cq to 256, and the weight of the loss λ1, λ2, λ3, and λ4 are set to 3, 1, 1.25, and 1, respectively. The rank r in Lo RA is set to 8. The model is trained with 60 epochs and the initial learning rate is reduced to 1e-5 at 50 iterations. ... All experiments are carried out on 4 RTX3090 GPUs with batch size set to 16. |