Deep-dense Conditional Random Fields for Object Co-segmentation
Authors: Zehuan Yuan, Tong Lu, Yirui Wu
IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our method on two datasets and achieve competitive performance. We train DDCRF network on PASCAL VOC 2012 dataset [Everingham et al., ] and test the performance on the widely used benchmark i Coseg [Batra et al., 2010] and one more challenging Internet dataset [Rubinstein et al., 2013]. |
| Researcher Affiliation | Academia | Zehuan Yuan1, Tong Lu1 , and Yirui Wu2 1State Key Laboratory for Novel Software Technology, Nanjing University, China 2College of Computer and Information, Hohai University, China |
| Pseudocode | No | The paper describes the model and training steps in narrative text and figures but does not provide structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain any statements or links regarding the public release of its source code. |
| Open Datasets | Yes | We train DDCRF network on PASCAL VOC 2012 dataset [Everingham et al., ] and test the performance on the widely used benchmark i Coseg [Batra et al., 2010] and one more challenging Internet dataset [Rubinstein et al., 2013]. |
| Dataset Splits | No | The paper describes the training data (PASCAL VOC 2012) and evaluation datasets (i Coseg, Internet), but it does not provide specific details on a separate validation dataset split used during model training or hyperparameter tuning. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run the experiments (e.g., GPU/CPU models, memory specifications). |
| Software Dependencies | No | The paper mentions deep learning architectures and techniques (VGG16, ROI Pooling, dense CRF), but it does not specify any software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | The entire network is fine-tuned on a pretrained model for Image Net classification except new layers. We randomly initialize all new layers by drawing weights from a zero-mean Gaussian distribution with the standard deviation 0.01. In order to share computations among object proposals, we follow the same sampling strategy as [Ren et al., 2015], namely, 2 images are sampled per iteration and 64 object proposals per image. Therefore, positives and negatives account for half respectively in each mini-batch with 128 proposals totally. In order to make optimization faster and converge to a good optima, a two-stage training is adopted to train the network. Firstly, we learn parameters without including pairwise term. ... with the initial learning rate 0.001. Stochastic Gradient Descent (SGD) is used as optimizer and the learning rate decreases to 0.0001 after 80K iterations. ... Additionally, we add dropout layers for all three branches. In the segmentation branch, fc8 is trained with the dropout rate 0.5. Two dropout layers with dropout rate 0.8 are followed after fc6 and fc7, respectively. |