Deep Mobile Matting

This is a lightweight image matting model in PyTorch.

Features

1. MobileNetV2 as backbone.
2. DeepLabv3 heads.
3. Small model (size: 23.5MB, FLOPs: 11.39GB, total params: 7.62 millions)

Performance

• The Composition-1k testing dataset.
• Evaluate with whole image.
• SAD normalized by 1000.
• Input image is normalized with mean=[0.485, 0.456, 0.406] and std=[0.229, 0.224, 0.225].
• Both erode and dialte to generate trimap.

Dependencies

• Python 3.6.8
• PyTorch 1.3

Dataset

Adobe Deep Image Matting Dataset

Follow the instruction to contact author for the dataset.

MSCOCO

Go to MSCOCO to download:

• 2014 Train images

PASCAL VOC

Go to PASCAL VOC to download:

• VOC challenge 2008 training/validation data
• The test data for the VOC2008 challenge

Usage

Data Pre-processing

Extract training images:

$ python pre_process.py
# python data_gen.py

Train

$ python train.py

If you want to visualize during training, run in your terminal:

$ tensorboard --logdir runs

Experimental results

The Composition-1k testing dataset

1. Test:

   $ python test.py

It prints out average SAD and MSE errors when finished.

Demo

Download pre-trained Deep Image Matting Link then run:

$ python demo.py

Image/Trimap | Output/GT | New BG/Compose |
|—-|—-|—-|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |