Recognizing Unseen Attribute-Object Pair with Generative Model
Authors: Zhixiong Nan, Yang Liu, Nanning Zheng, Song-Chun Zhu8811-8818
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conducted extensive experiments to compare our method with several state-of-the-art methods on two challenging datasets. The results show that our method outperforms all other methods. We also performed some ablation experiments to study the effect of individual loss function, the influence of visual feature extractor, and the interdependence of the attribute and object, from which we draw some important conclusions. |
| Researcher Affiliation | Academia | 1Xi an Jiaotong University, China 2University of California, Los Angeles, USA |
| Pseudocode | No | The paper describes its approach using text and mathematical equations, but does not include any explicit pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement about releasing its source code or a link to a code repository for the described methodology. |
| Open Datasets | Yes | Two datasets, MIT-States (Isola, Lim, and Adelson 2015) and UT-Zappos50K (Yu and Grauman 2014), are used for evaluation. |
| Dataset Splits | No | Following the same setting as in (Misra, Gupta, and Hebert 2017) and (Nagarajan and Grauman 2018), 1,262 attribute-object pairs are used for training and 700 pairs for testing. [...] We use 83 attribute-object pairs for training and 33 pairs for testing. The paper only mentions training and testing splits, not an explicit validation split. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. |
| Software Dependencies | No | We implement our end to end neural network with MXNet (Chen et al. 2015). The paper mentions MXNet but does not provide version numbers for it or any other software dependencies. |
| Experiment Setup | Yes | K in Eq. 9 is a parameter that controls the margin, and is set to 0.9 in our experiment. κ,α, β, γ in Eq. 12 are with the ratio of 1 : 0.2 : 2 : 2 for Mit-States dataset and 1 : 0.2 : 0.5 : 2 for UT-Zappos50K dataset. We use ADAM as our optimizer with the initial learning rate as 0.0001, which decays by 0.9 every two epochs. At every iteration we feed the minibatch to the network with the batch size as 128. |