Differentiable Meta-Learning Model for Few-Shot Semantic Segmentation
Authors: Pinzhuo Tian, Zhangkai Wu, Lei Qi, Lei Wang, Yinghuan Shi, Yang Gao12087-12094
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on two popular semantic segmentation datasets, i.e., PASCAL VOC and COCO, reveal the effectiveness of the proposed Meta Seg Net in the K-way few-shot semantic segmentation task. |
| Researcher Affiliation | Academia | 1National Key Laboratory for Novel Software Technology, Nanjing University, China 2School of Computing and Information Technology, University of Wollongong, Australia |
| Pseudocode | No | The paper describes the model architecture and its components but does not provide pseudocode or an algorithm block. |
| Open Source Code | No | The paper does not provide an explicit statement about the release of source code or a link to a code repository for the described methodology. |
| Open Datasets | Yes | We evaluate our framework on two semantic segmentation datasets, i.e., PASCAL VOC 2012 (Hariharan et al. 2015) and COCO 2014 (Lin et al. 2014), in the K-way, N-shot semantic segmentation task. |
| Dataset Splits | Yes | PASCAL VOC 2012 (Hariharan et al. 2015) contains 20 different object categories which consists of 10,582 and 1,449 images as the training and validation sets, respectively. (2) COCO 2014 (Lin et al. 2014) dataset is a challenging large-scale dataset which contains 80 different object categories. In COCO, 82,783 and 40,504 images are used for training and validation, respectively. |
| Hardware Specification | Yes | We implement our method by Py Torch on two NVIDIA 2080 Ti GPUs with 12 GB memory. |
| Software Dependencies | No | The paper mentions implementing the method using PyTorch and using Adam optimizer, but does not provide specific version numbers for these or other software dependencies. |
| Experiment Setup | Yes | For optimization, we use Adam (Kingma and Ba 2014) with learning rate as 0.001. For PASCAL, our model is meta-trained for 40 epochs, and each epoch consists of 1000 episodes. For COCO, our model is meta-trained for 80 epoches, and each epoch consists of 500 episodes. During meta-training, we adopt horizontal flip, and randomly rotate the image with 0, 90, 180 or 270 degree for data augmentation. |