pytorch-polygon-rnn

Pytorch implementation of Polygon-RNN. Notice that I use another way to handle the first vertex instead of training another model as in the paper.

Difference with the original paper

  1. I use two virtual starting vertices to handle the first vertex as in the image captioning.

  2. I add a LSTM layer after the ConvLSTM layers since I need the output to be D*D+1 dimension to handle the end symbol.

How to train and test

  1. Download data from CityScapes, organize the image files and annotation json files as follows:
img
ā”œā”€ā”€ train
ā”‚   ā”œā”€ā”€ cityname1
ā”‚   ā”‚   ā”œā”€ā”€ pic.png
ā”‚   ā”‚   ā”œā”€ā”€ ...
ā”‚   ā”œā”€ā”€ cityname2
ā”‚   ā”‚   ā”œā”€ā”€ pic.png
ā”‚   ā”‚   ā”œā”€ā”€ ...
ā”œā”€ā”€ val
ā”‚   ā”œā”€ā”€ cityname
ā”‚   ā”‚   ā”œā”€ā”€ pic.png
ā”‚   ā”‚   ā”œā”€ā”€ ...
ā”œā”€ā”€ test
ā”‚   ā”œā”€ā”€ cityname
ā”‚   ā”‚   ā”œā”€ā”€ pic.png
ā”‚   ā”‚   ā”œā”€ā”€ ...
label
ā”œā”€ā”€ train
ā”‚   ā”œā”€ā”€ cityname1
ā”‚   ā”‚   ā”œā”€ā”€ annotation.json
ā”‚   ā”‚   ā”œā”€ā”€ ...
ā”‚   ā”œā”€ā”€ cityname2
ā”‚   ā”‚   ā”œā”€ā”€ annotation.json
ā”‚   ā”‚   ā”œā”€ā”€ ...
ā”œā”€ā”€ val
ā”‚   ā”œā”€ā”€ cityname
ā”‚   ā”‚   ā”œā”€ā”€ annotation.json
ā”‚   ā”‚   ā”œā”€ā”€ ...
ā”œā”€ā”€ test
ā”‚   ā”œā”€ā”€ cityname
ā”‚   ā”‚   ā”œā”€ā”€ annotation.json
ā”‚   ā”‚   ā”œā”€ā”€ ...

The png files and the json files should have corresponding same name.

Execute the following command to make directories for new data and save models:

mkdir -p new_img/(train/val/test)
mkdir -p new_label/(train/val/test)
mkdir save
  1. Run the following command to generate data for train/validation/test.

    python generate_data.py --data train/val/test
  2. Run the following command to train.

    python train.py --gpu_id 0 1 2 --batch_size 8  --lr 0.0001
  3. Run the following command to validate.

    python validate.py --gpu_id 0 1 2 --batch_size 8
  4. Run the following command to test.

    python test.py --gpu_id 0 --model ./save/model.pth

Now you can easily change configurations on default_config.yaml.

Pytorch Polygon Rnn

Pytorch implementation of Polygon-RNN(http://www.cs.toronto.edu/polyrnn/poly_cvpr17/)

Pytorch Polygon Rnn Info

ā­ Stars 84
šŸ”— Source Code github.com
šŸ•’ Last Update a year ago
šŸ•’ Created 3 years ago
šŸž Open Issues 4
āž— Star-Issue Ratio 21
šŸ˜Ž Author AlexMa011