Stamatis Alexandropoulos, Christos Sakaridis (ETH), Petros Maragos (NTUA)

Semantic segmentation is a fundamental task in visual scene understanding. We focus on the supervised setting, where ground-truth semantic annotations are available. Based on knowledge about the high regularity of real-world scenes, we propose a method for improving class predictions by learning to selectively exploit information from neighboring pixels. In particular, our method is based on the prior that for each pixel, there is a seed pixel in its close neighborhood sharing the same prediction with the former. Motivated by this prior, we design a novel two-head network, named Offset Vector Network (OVeNet), which generates both standard semantic predictions and a dense 2D offset vector field indicating the offset from each pixel to the respective seed pixel, which is used to compute an alternative, seed-based semantic prediction. The two predictions are adaptively fused at each pixel using a learnt dense confidence map for the predicted offset vector field. We supervise offset vectors indirectly via optimizing the seed-based prediction and via a novel loss on the confidence map. Compared to the baseline state-of-the-art architectures HRNet and HRNet+OCR on which OVeNet is built, the latter achieves significant performance gains on two prominent benchmarks for semantic segmentation, namely Cityscapes, ACDC and ADE20K.

arXiv (pdf)

This is the reference PyTorch implementation for training and evaluation of HRNet using the method described in this paper.

This software is released under a creative commons license which allows for personal and research use only. For a commercial license please contact the authors. You can view a license summary here.

OVeNet consists of two output heads. The first head outputs pixel-level Logits (C), while the second head outputs a dense offset vector field (o) identifying positions of seed pixels along with a confidence map (F). Then, the coefficients of seed pixels are used to predict classes at each position. The resulting prediction (S_s) is adaptively fused with the initial prediction (S_i) using the confidence map F to compute the final prediction S_f

1. Installation
2. Training
3. Evaluation
4. Citation
5. Contributions

For setup, you need:

1. Linux
2. NVIDIA GPU with CUDA & CuDNN
3. Python 3
4. Conda
5. PyTorch=1.1.0 following the official instructions
6. Install dependencies: pip install -r requirements.txt

You need to download the Cityscapes, ACDC and ADE20K datasets.

Your directory tree should be look like this:

$ROOT/data
├── cityscapes
│   ├── gtFine
│   │   ├── test
│   │   ├── train
│   │   └── val
│   └── leftImg8bit
│       ├── test
│       ├── train
│       └── val
├── acdc
│   ├── gt
│   │   ├── fog
│   │   ├── night
│   │   └── rain
│   │   └── snow
│   └── rgb_anon
│   │   ├── fog
│   │   ├── night
│   │   └── rain
│   │   └── snow
├── ade20k
│   ├── train
│   │   ├── image
│   │   └── label
│   └── val
│       ├── image
│       └── label
├── list
│   ├── cityscapes
│   │   ├── test.lst
│   │   ├── trainval.lst
│   │   └── val.lst
│   ├── acdc
│   │   ├── test.lst
│   │   ├── trainval.lst
│   │   └── val.lst

Please specify the configuration file.

The models are initialized by the weights pretrained on the ImageNet. You can download the pretrained models from here. All the others pretrained models can be found here.

For example, train the HRNet-W48 on Cityscapes with a batch size of 8 on 4 GPUs:

python -m torch.distributed.launch --nproc_per_node=4 tools/train.py --cfg  experiments/cityscapes/seg_hrnet_w48_train_ohem_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484_cityscapes_pretrained.yaml

For example, evaluating our model on the Cityscapes validation set with multi-scale and flip testing:

python tools/test.py --cfg experiments/cityscapes/seg_hrnet_w48_train_ohem_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484_cityscapes_pretrained.yaml \
                     TEST.MODEL_FILE output/cityscapes/seg_hrnet_w48_train_ohem_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484_cityscapes_pretrained_base_experiment/final_state.pth \
                     TEST.SCALE_LIST 0.5,0.75,1.0,1.25,1.5,1.75 \
                     TEST.FLIP_TEST True

Evaluating our model on the Cityscapes test set with multi-scale and flip testing:

python tools/test.py --cfg experiments/cityscapes/seg_hrnet_w48_train_ohem_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484_cityscapes_pretrained.yaml \
                     DATASET.TEST_SET list/cityscapes/test.lst \
                     TEST.MODEL_FILE output/cityscapes/seg_hrnet_w48_train_ohem_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484_cityscapes_pretrained_base_experiment/final_state.pth \
                     TEST.SCALE_LIST 0.5,0.75,1.0,1.25,1.5,1.75 \
                     TEST.FLIP_TEST True

If you find our work useful in your research please use this identifier to cite or link to this item.

If you find any bug in the code. Please report to
Stamatis Alexandropoulos ([email protected])