Implementation of YOLO v1 object detector in PyTorch. Full tutorial can be found here in korean.
Tested under Python 3.6, PyTorch 0.4.1 on Ubuntu 16.04, Windows10.
See requirements for details.
NOTICE: different versions of PyTorch package have different memory usages.
main.py --mode train -data_path where/your/dataset/is --class_path ./names/VOC.names --num_class 20 --use_augmentation True --use_visdom True
main.py --mode test --data_path where/your/dataset/is --class_path ./names/VOC.names --num_class 20 --checkpoint_path your_checkpoint.pth.tar
python3 utilities/download_checkpoint.pyOnly Pascal VOC datasets are supported for now.
| argument | type | description | default |
|---|---|---|---|
| --mode | str | train or test | train |
| --dataset | str | only support voc now | voc |
| --data_path | str | data path | |
| --class_path | str | filenames text file path | |
| --input_height | int | input height | 448 |
| --input_width | int | input width | 448 |
| --batch_size | int | batch size | 16 |
| --num_epochs | int | # of epochs | 16000 |
| --learning_rate | float | initial learning rate | 1e-3 |
| --dropout | float | dropout probability | 0.5 |
| --num_gpus | int | # of GPUs for training | 1 |
| --checkpoint_path | str | checkpoint path | ./ |
| --use_augmentation | bool | image Augmentation | True |
| --use_visdom | bool | visdom | False |
| --use_wandb | bool | wandb | False |
| --use_summary | bool | descripte Model summary | True |
| --use_gtcheck | bool | gt check flag | False |
| --use_githash | bool | use githash | False |
| --num_class | int | number of classes | 5 |
This project is equally contributed by Chanhee Jeong, Donghyeon Hwang, and Jaewon Lee.
See LICENSE for details.
[1] Redmon, Joseph, et al. "You only look once: Unified, real-time object detection." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016. (https://arxiv.org/abs/1506.02640)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent