GTNet: Generative Transfer Network for Zero-Shot Object Detection
Authors: Shizhen Zhao, Changxin Gao, Yuanjie Shao, Lerenhan Li, Changqian Yu, Zhong Ji, Nong Sang12967-12974
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our method on three public datasets and the results demonstrate that our method performs favorably against the state-of-the-art ZSD approaches. We conduct extensive experiments on three public datasets, and those experimental results show that the proposed approach performs favorably against the state-of-the-art ZSD algorithms. |
| Researcher Affiliation | Academia | Shizhen Zhao,1 Changxin Gao,1 Yuanjie Shao,1 Lerenhan Li,1 Changqian Yu,1 Zhong Ji,2 Nong Sang1 1Key Laboratory of Ministry of Education for Image Processing and Intelligent Control, School of Artificial Intelligence and Automation, Huazhong University of Science and Technology 2School of Electrical and Information Engineering, Tianjin University Email: {zhaosz, cgao}@hust.edu.cn |
| Pseudocode | No | The paper describes the methods and processes using text and mathematical equations, but it does not contain any formal pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code, nor does it include any links to a code repository. |
| Open Datasets | Yes | Our proposed framework is evaluated on three challenging datasets. ILSVRC-2017 detection dataset (Russakovsky et al. 2015) contains 200 object categories. ... MSCOCO (Lin et al. 2014) was introduced for object detection and semantic segmentation tasks... Visual Genome (VG) (Krishna et al. 2017) was collected with a focus on visual relationship understanding. |
| Dataset Splits | Yes | For MSCOCO, following (Bansal et al. 2018), training images are selected from the 2014 training set and testing images are randomly sampled from the validation set. The test set is composed of the rest of images from ILSVRC training dataset and images from validation dataset that have at least one unseen class bounding box. |
| Hardware Specification | No | The paper mentions using Resnet-101 as the backbone for the feature extractor and notes that modules are "trained using the Adam optimizer," but it does not provide specific details about the hardware (e.g., GPU models, CPU types, memory) used for these experiments. |
| Software Dependencies | No | The paper mentions using the Adam optimizer and models like Faster R-CNN and WGAN, but it does not provide specific version numbers for any software dependencies or libraries used in the implementation. |
| Experiment Setup | Yes | We use β1 = β2 = β3 = 0.01 and γ1 = γ2 = γ3 = 0.1 accross all datasets. Following (Bansal et al. 2018), the parameters tf and tb are set to 0.5 and 0.2. All the modules are trained using the Adam optimizer with a 10 4 learning rate. |