Skip to content

Latest commit

 

History

History
160 lines (109 loc) · 4.72 KB

GETTING_STARTED.md

File metadata and controls

160 lines (109 loc) · 4.72 KB

Getting Started

This page provides basic tutorials about the usage of mmdetection. For installation instructions, please see INSTALL.md.

Prepare ROAS dataset.

It is recommended to symlink the dataset root to roas/data.

Here, we give an example for single scale data preparation of ROAS dataset.

First, make sure your initial data are in the following structure.

data/roas
├── train
│   ├── images
│   └── json
└── test 
    ├── images
    └── json

Split the original images and create COCO format json.

python ROAS_devkit/prepare_roas.py --srcpath path_to_dota --dstpath path_to_split_data

Then you will get data in the following structure

roas16-split
├── test1024_2x
│   ├── coco.json
│   └── images
└── train768_2x
     ├── coco.json
     └── images

Inference with pretrained models

Test a dataset

  • single GPU testing
  • multiple GPU testing

You can use the following commands to test a dataset.

# single-gpu testing
python tools/test.py ${CONFIG_FILE} ${CHECKPOINT_FILE} [--out ${RESULT_FILE}]

# multi-gpu testing
./tools/dist_test.sh ${CONFIG_FILE} ${CHECKPOINT_FILE} ${GPU_NUM} [--out ${RESULT_FILE}]

Optional arguments:

  • RESULT_FILE: Filename of the output results in pickle format. If not specified, the results will not be saved to a file.

Examples:

Assume that you have already downloaded the checkpoints to work_dirs/.

  1. Test Faster R-CNN with RoI Transformer. Pretrained Weight: RoI Transformer with Resnet101
python tools/test.py configs/ROAS/faster_rcnn_RoITrans_r101_fpn_2x_roas16.py \
    work_dirs/faster_rcnn_RoITrans_r101_fpn_2x_roas16/epoch_12.pth
  1. Test Faster R-CNN with RoI Ttransformer with 4 GPUs.
./tools/dist_test.sh configs/ROAS/faster_rcnn_RoITrans_r101_fpn_2x_roas16.py \
    work_dirs/faster_rcnn_RoITrans_r101_fpn_2x_roas16/epoch_12.pth  \
    4
  1. Parse the results.pkl to the format which is used in DOTA evaluation

For methods with only OBB Head, set the type OBB.

python tools/parse_results.py --config configs/ROAS/faster_rcnn_RoITrans_r101_fpn_2x_roas16.py --type OBB
  1. Merge results as csv
python3 ROAS_devkit/merge_results_as_csv.py --srcpath ${PARSING PATH}$ --dstpath ${DST_PATH}
  1. Evaluate performance
python3 ROAS_devkit/evaluate_roas.py --gt_csv_path ${GT_PATH}$ --pred_csv_path ${PRED_PATH}

You have to prepare csv file for ground-truth by ./ROAS_devkit/geojson2csv.py.

Demo of inference in a large size image.

python demo_large_image.py

Train a model

mmdetection implements distributed training and non-distributed training, which uses MMDistributedDataParallel and MMDataParallel respectively.

All outputs (log files and checkpoints) will be saved to the working directory, which is specified by work_dir in the config file.

*Important*: The default learning rate in config files is for 8 GPUs. If you use less or more than 8 GPUs, you need to set the learning rate proportional to the GPU num, e.g., 0.01 for 4 GPUs and 0.04 for 16 GPUs.

Train with a single GPU

python tools/train.py ${CONFIG_FILE}

If you want to specify the working directory in the command, you can add an argument --work_dir ${YOUR_WORK_DIR}.

Train with multiple GPUs

./tools/dist_train.sh ${CONFIG_FILE} ${GPU_NUM} [optional arguments]

Optional arguments are:

  • --validate (recommended): Perform evaluation at every k (default=1) epochs during the training.
  • --work_dir ${WORK_DIR}: Override the working directory specified in the config file.
  • --resume_from ${CHECKPOINT_FILE}: Resume from a previous checkpoint file.

Train with multiple machines

If you run mmdetection on a cluster managed with slurm, you can just use the script slurm_train.sh.

./tools/slurm_train.sh ${PARTITION} ${JOB_NAME} ${CONFIG_FILE} ${WORK_DIR} [${GPUS}]

Here is an example of using 16 GPUs to train Mask R-CNN on the dev partition.

./tools/slurm_train.sh dev mask_r50_1x configs/mask_rcnn_r50_fpn_1x.py /nfs/xxxx/mask_rcnn_r50_fpn_1x 16

You can check slurm_train.sh for full arguments and environment variables.

If you have just multiple machines connected with ethernet, you can refer to pytorch launch utility. Usually it is slow if you do not have high speed networking like infiniband.

For more information on how it works, you can refer to TECHNICAL_DETAILS.md (TODO).