Dense Transformer Networks for Brain Electron Microscopy Image Segmentation
Authors: Jun Li, Yongjun Chen, Lei Cai, Ian Davidson, Shuiwang Ji
IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We apply the proposed networks on biological image segmentation tasks and show superior performance is achieved in comparison to baseline methods. Experimental results on biological images demonstrate the effectiveness of the proposed dense transformer networks. We evaluate the proposed methods on two image segmentation tasks. The U-Net [Ronneberger et al., 2015] is adopted as our base model in both tasks, as it has achieved state-of-the-art performance on image segmentation tasks. The experimental results in terms of ROC curves are provided in Figure 4. We can observe that the proposed DTN model achieves higher performance than the baseline U-Net model, improving AUC from 0.8676 to 0.8953. Table 1 shows the comparison of training and prediction time between the U-Net model and the proposed DTN model on the two data sets. |
| Researcher Affiliation | Academia | 1Washington State University 2University of California, Davis 3Texas A&M University {jun.li3, yongjun.chen, lei.cai}@wsu.edu, davidson@cs.ucdavis.edu, sji@tamu.edu |
| Pseudocode | No | The paper describes mathematical formulations and processes but does not include any pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any statement or link regarding the public availability of its source code. |
| Open Datasets | Yes | We use data set from the 3D Segmentation of Neurites in EM Images (SNEMI3D, http://brainiac2.mit.edu/SNEMI3D/). The SNEMI3D data set consists of 100 1024 1024 EM image slices. |
| Dataset Splits | No | The paper mentions using "224x224 patches from the original images and use it to train the networks" but does not specify a validation set or explicit train/validation/test splits needed for reproduction. |
| Hardware Specification | Yes | Table 1: Training and prediction time on the two data sets using a Tesla K40 GPU. |
| Software Dependencies | No | The paper mentions the use of "deep convolution neural networks (CNNs)" and refers to architectures like "U-Net" but does not specify any software names with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | The network consists of 5 layers in the encoder path and another corresponding 5 layers in the decoder path. A stack of two 3x3 convolutional layers have the same receptive field as a 5x5 convolutional layer, but with less parameters. Therefore, we use 3x3 kernels and one pixel padding to retain the size of feature maps at each level. We insert the spatial encoder layer and dense transformer decoder layer into corresponding positions at the same level. The layers are applied to the 4th layer. As for the transformation layers, we use 16 fiducial points that are evenly distributed on the output feature maps. We randomly pick 224x224 patches from the original images and use it to train the networks. We compare the training time of 10,000 iterations and prediction time of 40 (SNEMI3D) images. |