[CVPR‘20] SpixelFCN: Superpixel Segmentation with Fully Convolutional Network
This is is a PyTorch implementation of the superpixel segmentation network introduced in our CVPR-20 paper:
Please contact Fengting Yang ([email protected]) if you have any question.
The training code was mainly developed and tested with python 2.7, PyTorch 0.4.1, CUDA 9, and Ubuntu 16.04.
During test, we make use of the component connection method in SSN to enforce the connectivity in superpixels. The code has been included in
/third_paty/cython. To compile it:
cd third_party/cython/ python setup.py install --user cd ../..
The demo script
run_demo.pyprovides the superpixels with grid size
16 x 16using our pre-trained model (in
/pretrained_ckpt). Please feel free to provide your own images by copying them into
/demo/inputs, and run
python run_demo.py --data_dir=./demo/inputs --data_suffix=jpg --output=./demoThe results will be generate in a new folder under
To generate training and test dataset, please first download the data from the original BSDS500 dataset, and extract it to. Then, run
cd data_preprocessing python pre_process_bsd500.py --dataset= --dump_root= python pre_process_bsd500_ori_sz.py --dataset= --dump_root= cd ..The code will generate three folders under the , named as
/test, and three
.txtfiles record the absolute path of the images, named as
Once the data is prepared, we should be able to train the model by running the following command
python main.py --data= --savepath=
if we wish to continue a train process or fine-tune from a pre-trained model, we can run
python main.py --data= --savepath= --pretrained=The code will start from the recorded status, which includes the optimizer status and epoch number.
The training log can be viewed from the
tensorboardsession by running
tensorboard --logdir= --port=8888
If everything is set up properly, reasonable segmentation should be observed after 10 epochs.
We provide test code to generate: 1) superpixel visualization and 2) the
.csvfiles for evaluation.
To test on BSDS500, run
python run_infer_bsds.py --data_dir= --output= --pretrained=
To test on NYUv2, please first extract our pre-processed dataset from
/nyu_test_set/nyu_preprocess_tst.tar.gzto , or follow the intruction on the superpixel benchmark to generate the test dataset, and then run
python run_infer_nyu.py --data_dir= --output= --pretrained=
To test on other datasets, please first collect all the images into one folder, and then convert them into the same format (e.g.
.jpg) if necessary, and run
python run_demo.py --data_dir= --data_suffix= --output= --pretrained=Superpixels with grid size
16 x 16will be generated by default. To generate the superpixel with a different grid size, we simply need to resize the images into the approporate resolution before passing them through the code. Please refer to
run_infer_nyu.pyfor the details.
(1) download the code and build it accordingly;
(2) edit the variables
cp /eval_spixel/my_eval.sh /examples/bash/ cd /examples/ bash my_eval.shseveral files should be generated in the
map_csvfolders in the corresponding test outputs;
cd eval_spixel python copy_resCSV.py --src= --dst=(5) open
/eval_spixel/plot_benchmark_curve.m, set the
our1l_res_pathas and modify the
num_listaccording to the test setting. The default setting is for our BSDS500 test set.;
(6) run the
CO Score, and
BR-BP curveof our method should be shown on the screen. If you wish to compare our method with the others, you can first run the method and organize the data as we state above, and uncomment the code in the
plot_benchmark_curve.mto generate a similar figure shown in our papers.
Our code is developed based on the training framework provided by FlowNetPytorch.