https://github.com/jahongir7174/dualhtc
DualHTC++ implementation using PyTorch (supports MOSAIC/MixUp and RandomAugment)
https://github.com/jahongir7174/dualhtc
cbnetv2 htc hybrid-task-cascade instance-segmentation mixup mosaic object-detection pytorch
Last synced: about 1 month ago
JSON representation
DualHTC++ implementation using PyTorch (supports MOSAIC/MixUp and RandomAugment)
- Host: GitHub
- URL: https://github.com/jahongir7174/dualhtc
- Owner: jahongir7174
- License: mit
- Created: 2021-01-14T08:27:27.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2022-03-15T00:15:51.000Z (about 4 years ago)
- Last Synced: 2025-12-28T16:24:04.429Z (5 months ago)
- Topics: cbnetv2, htc, hybrid-task-cascade, instance-segmentation, mixup, mosaic, object-detection, pytorch
- Language: Python
- Homepage:
- Size: 31.3 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
[DualHTC++](https://arxiv.org/abs/2107.00420) implementation using PyTorch (supports `MOSAIC` and `MixUp`)
#### Train
* Install [mmdet](https://github.com/open-mmlab/mmdetection) toolbox
* Register your dataset to `utils/dataset.py`, see `INSDataset`
* See `nets/exp01.py` for using `MOSAIC` and `MixUp` data pipeline
* See `nets/exp02.py` for using `RandomAugment` data pipeline
* Run `bash ./main.sh ./nets/exp01.py $ --train` for training, `$` is number of GPUs
#### Note
* The default configuration is for [2021 VIPriors Instance Segmentation Challenge](https://competitions.codalab.org/competitions/33340) dataset
#### Reference
* https://github.com/ultralytics/yolov5
* https://github.com/open-mmlab/mmdetection