Skip to content

Latest commit

 

History

History
98 lines (74 loc) · 3.79 KB

README.md

File metadata and controls

98 lines (74 loc) · 3.79 KB
Raw

Data Preparation of MS COCO

Language: 简体中文 | English

There are 5 python files in this folder, which are used for converting MS COCO JSON annotation format to PASCAL VOC XML annotation format (refer to KapilM26/coco2VOC).

The following instructions are for data preparation of MS COCO. Please ensure you have finished the installation in installation.md.

  1. Please create a file directory tree as below, where the images and annotations of MS COCO ( train and val ) can be downloaded from here:

    Images ( train ): http://images.cocodataset.org/zips/train2017.zip

    Images ( val ): http://images.cocodataset.org/zips/val2017.zip

    Annotations: http://images.cocodataset.org/annotations/annotations_trainval2017.zip

    ├── coco (download from official website)
│   ├── annotations
│   │   ├── instances_train2017.json
│   │   ├── instances_val2017.json
│   ├── images (combination of train2017 and val2017)
│   ├── train2017
│   ├── val2017
├── coco2voc (manually create)
│   ├── Annotations (empty)
│   ├── ImageSets
│   │   ├── Main (empty)
│   ├── JPEGImages (symbol link -> coco/images)

    You may also use the following commands directly:

    cd $YOUR_DATASET_PATH
mkdir coco
cd coco
wget http://images.cocodataset.org/zips/train2017.zip  # it would cost much time
unzip train2017.zip  # it would cost much time
wget http://images.cocodataset.org/zips/val2017.zip
unzip val2017.zip
wget http://images.cocodataset.org/annotations/annotations_trainval2017.zip
unzip annotations_trainval2017.zip
mkdir images
cp train2017/* images/
cp val2017/* images/
cd ../
mkdir coco2voc
mkdir cocovoc/Annotations
mkdir coco2voc/ImageSets
mkdir coco2voc/ImageSets/Main
ln -s coco/images coco2voc/JPEGImages

  2. Please modify the corresponding dataset directory in MIAOD.py in this folder, they are located in:

    Line 1: data_root_coco='$YOUR_DATASET_PATH/coco/'
Line 2: data_root_voc='$YOUR_DATASET_PATH/coco2voc/'

    Please change the $YOUR_DATASET_PATHs above to your actual dataset directory (i.e., the directory where you create the coco folder).

    And please use the absolute path (i.e., start with /) but not a relative path (i.e., start with ./ or ../).

  3. Please copy these 4 files to the following directory and replace the corresponding original files:

    • active_datasets.py -> mmdet/utils/active_datasets.py
    • MIAOD.py -> configs/MIAOD.py
    • train.py -> tools/train.py
    • voc.py -> mmdet/datasets/voc.py

    The commands are:

    cp active_datasets.py ../mmdet/utils/active_datasets.py
cp MIAOD.py ../configs/MIAOD.py
cp train.py ../tools/train.py
cp voc.py ../mmdet/datasets/voc.py

  4. Please install pascal_voc_writer package to write PASCAL VOC XML annotation as below.

    pip install pascal_voc_writer

  5. Please convert the MS COCO JSON annotation format to PASCAL VOC XML annotation format as below.

    python coco2voc.py \
    --ann_file $YOUR_DATASET_PATH/coco/annotations/instances_train2017.json \
    --output_dir $YOUR_DATASET_PATH/coco2voc/Annotations/

    Please change the $YOUR_DATASET_PATHs above to your actual dataset directory (i.e., the directory where you create the coco folder).

Now you have finished the data preparation on MS COCO. Please follow here for remaining training and test steps, and please replace --eval mAP with --eval bbox in the test step for MS COCO.