CSAILVision ADE20K

show more

Implementations

Semantic Segmentation on MIT ADE20K dataset in PyTorch.

PyTorch implementation of semantic segmentation models on MIT ADE20K scene parsing dataset.

Encoder:

  • MobileNetV2dilated
  • ResNet18dilated
  • ResNet50dilated
  • ResNet101dilated

Decoder:

  • C1 (1 convolution module)
  • C1_deepsup (C1 + deep supervision trick)
  • PPM (Pyramid Pooling Module, see PSPNet paper for details.)
  • PPM_deepsup (PPM + deep supervision trick)
  • UPerNet (Pyramid Pooling + FPN head, see UperNet for details.)


The code is developed under the following configurations.

  • Hardware: >=4 GPUs for training, >=1 GPU for testing (set [--gpus GPUS] accordingly)
  • Software: Ubuntu 16.04.3 LTS, CUDA>=8.0, Python>=3.5, PyTorch>=0.4.0
  • Dependencies: numpy, scipy, opencv, yacs, tqdm

Github: https://github.com/CSAILVision/semantic-segmentation-pytorch

pytorchade20k
How helpful was this page?