Skip to content

OpenVINO™ Training Extensions - instance_segmentation - pytorch #98

@junxnone

Description

@junxnone

Reference

Brief

segmentoly
├── data
│   ├── dataparallel.py ==> 多 gpu training data process
│   └── transforms.py ==> augmentation 相关内容
├── datasets
│   ├── coco.py ==> define coco dataset
│   ├── factory.py  ==> get coco dataset [train2017/val2017]
│   ├── images.py  ==> return image dataset dict
│   ├── instance_dataset.py ==> define InstanceDataset
│   └── video.py ==> define VideoDataset
├── extensions
│   ├── deformable_conv ==> deformable conv cpu/gpu 实现
│   ├── nms ==> nms cpu/gpu 实现
│   │   ├── cpu
│   │   └── gpu
│   └── roi_align ==>roi align cpu/gpu 实现
│       ├── cpu
│       └── gpu
├── rcnn
│   ├── backbones
│   ├── model_zoo
└── utils

tools/
├── convert_to_onnx.py
├── demo.py
├── download_pretrained_weights.py
├── finetune_0050.py
├── model_zoo.yaml
├── test.py
├── train_0050.py
└── train.py

Augmentation

  • Compose(object) - 串行做 augmentation
Function Description
Resize(object)
RandomResize(object)
RandomHorizontalFlip(object)
FilterByBoxSize(object)
ToTensor(object)
Normalize(object)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions