Deep Propagation Based Image Matting
Authors: Yu Wang, Yi Niu, Peiyong Duan, Jianwei Lin, Yuanjie Zheng
IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our framework results in a semantic-level pairwise similarity of pixels for propagation by learning deep image representations adapted to matte propagation. It combines the power of deep learning and matte propagation and can therefore surpass prior state-of-the-art matting techniques in terms of both accuracy and training complexity, as validated by our experimental results from 243K images created based on two benchmark matting databases. |
| Researcher Affiliation | Academia | Yu Wang1, Yi Niu1, Peiyong Duan1, , Jianwei Lin1, Yuanjie Zheng1,2,3,4, 1 School of Information Science and Engineering, Shandong Normal University, China 2 Key Lab of Intelligent Computing and Information Security in Universities of Shandong, China 3 Shandong Provincial Key Lab for Distributed Computer Software Novel Technology, China 4 Institute of Biomedical Sciences, Shandong Normal University, China {wangyu52, linjianwei}@stu.sdnu.edu.cn, {niuyi, duanpeiyong, yjzheng}@sdnu.edu.cn |
| Pseudocode | No | The paper describes the model architecture and mathematical formulations but does not include any pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | The paper does not provide any concrete access information (e.g., repository link, explicit statement of release, or mention in supplementary materials) for the source code of their methodology. |
| Open Datasets | Yes | We evaluate the performance of the proposed Deep Matte Prop Net on two matting tasks, the benchmark alphamatting.com dataset [Rhemann et al., 2009] and our own dataset. As the benchmark challenge for image matting, the alphamatting.com dataset makes both the original images and ground truth mattes available online (at www.alphamatting.com). |
| Dataset Splits | No | The paper states: 'We treat the 50 images including the 8 alphamatting.com testing images plus 42 images selected randomly from other 73 original images and all their composited images as the training set and all the left images as the testing set.' It does not explicitly mention a validation set or its split details. |
| Hardware Specification | Yes | We implement the Deep Matte Prop Net using Caffe and conduct training and testing on a NVIDIA Titan X graphics card. |
| Software Dependencies | No | The paper states 'We implement the Deep Matte Prop Net using Caffe' but does not specify a version number for Caffe or any other software dependencies. |
| Experiment Setup | Yes | The training is carried out with a fixed learning rate of 0.1 and momentum of 0.9. The overall training phase requires about 20 epochs. |