hardyqr / PyTorch-Segmentation-Zoo

A PyTorch collection of semantic segmentation tools.

PyTorch Semantic Segmentation Zoo

• Currently under construction.

run

python3 train.py --train_rgb_path ./data/image/train \
                 --train_depth_path ./data/depth/train \
                 --train_label_path ./data/label37/train \
                 --val_rgb_path ./data/image/test \
                 --val_depth_path ./data/depth/test \
                 --val_label_path ./data/label37/test \
                 --batch_size 2 \
                 --lr 0.001 \
                 --n_classes 38 \
                 --model Deeplab-v2 \
                 --data SUNRGBD \
                 --log_path runs/test \
                 --epochs 30

Requirements

• PyTorch >= 4.0
• tensorboardX

Supported models

(Most models are referenced from open source projects. Please check top of [model].py for source(s).)

• U-Net
• ResNet-DUC-HDC
• DeepLab-v2 (ResNet)
• DeepLab-LargeFOV (VGG16)

Supported datasets

• SUN RGB-D
• Pascal VOC 2012

TODO

• tensorboardX for visualizing logs
• inference.py