Lots of Image Semantic Segmentation Implementations in Tensorflow/Keras
Highly inspired by HERE
Currently, under-development :(
- python 3.x
- tensorflow 1.x
- keras 2.x
- numpy
- scikit-image
- opencv-python
- h5py
- tqdm
$ sudo python3 -m pip install -r requirements.txt
(Before running train.py, MAKE SURE run after downloading DataSet & changing DataSet's directory in xxx_train.py)
just after it, RUN train.py
$ python3 xxx_train.py
- FCNet
- SegNet
- U-Net
- FusionNet
- FC-DenseNet
- ENet
- LinkNet
- RefineNet
- PSPNet
- Mask R-CNN
- DecoupledNet
- GAN-SS
- G-FRNet
MS COCO 2017DataSet will be used!
| DataSet | Train | Validate | Test | Disk |
|---|---|---|---|---|
| MS COCO 2017 | 118287 | 5000 | 40670 | 26.3GB |
│
├── xxNet
│ ├──gan_img (generated images)
│ │ ├── train_xxx.png
│ │ └── train_xxx.png
│ ├── model (model)
│ │ └── model.txt (google-drive link for pre-trained model)
│ ├── xxx_model.py (model)
│ ├── xxx_train.py (trainer)
│ ├── xxx_tb.png (Tensor-Board result)
│ └── readme.md (results & explains)
├── metrics.py (metrics)
├── tfutil.py (useful TF utils)
├── image_utils.py (image processing)
└── datasets.py (DataSet loader)
Here's a google drive link. You can download pre-trained models from here !
| Name | Summary | Paper | Code |
|---|---|---|---|
| FCN | Fully Convolutional Networks for Semantic Segmentation | [arXiv] | |
| SegNet | A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation | [arXiv] | |
| U-Net | Convolutional Networks for Biomedical Image Segmentation | [arXiv] | |
| FusionNet | A deep fully residual convolutional neural network for image segmentation in connectomics | [arXiv] | |
| FC-DenseNet | Fully Convolutional DenseNets for Semantic Segmentation | [arXiv] | |
| ENet | A Deep Neural Network Architecture for Real-Time Semantic Segmentation | [arXiv] | |
| LinkNet | Exploiting Encoder Representations for Efficient Semantic Segmentation | [arXiv] | |
| Mask R-CNN | Mask R-CNN | [arXiv] | |
| PSPNet | Pyramid Scene Parsing Network | [arXiv] | |
| RefineNet | Multi-Path Refinement Networks for High-Resolution Semantic Segmentation | [arXiv] | |
| G-FRNet | Gated Feedback Refinement Network for Dense Image Labeling | [CVPR2017] | |
| DeepLabv3+ | Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation | [arXiv] | |
| DecoupledNet | Decoupled Deep Neural Network for Semi-supervised Semantic Segmentation | [arXiv] | |
| GAN-SS | Semi and Weakly Supervised Semantic Segmentation Using Generative Adversarial Network | [arXiv] |
- Implement FCN
- Implement Mask R-CNN
- Upload U-Net (Tuned)
- Upload FC-DenseNet
- Upload DeepLabv3+
Any suggestions and PRs and issues are WELCONE :)
HyeongChan Kim / @kozistr
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent