Pose2COCO Converter

Reproduction Guide for Converting OpenPose Annotations to COCO Format

---

Directory Structure

Ensure your directories are structured as follows before running the scripts:

Project_Root/
│
├── OpenPose_JSONs/                     # Directory containing all OpenPose JSON files
│   ├── file1.json
│   ├── file2.json
│   └── ... (all your OpenPose output JSONs)
│
├── Template/
│   └── NFD_root-15.json                # The base COCO JSON annotation template
│
├── Remapped_COCO_Output/               # Directory where the final COCO JSON file will be saved
│   └── NFD_root-15.json                # (Will be overwritten by the script)
│
└── Scripts/
    ├── json_copy.py                    # Script for remapping and converting annotations
    └── file_handler.py                 # Script for managing and cleaning files

Step-by-Step Instructions to Run the Scripts

1. Set Up Your Environment:

   • Make sure you have Python 3.x installed on your system.
   • Ensure that json and os Python libraries are available (these are part of the Python standard library, so no need for extra installations).

2. Generate OpenPose Annotations:

   • Use OpenPose to annotate your images. Save the output JSON files in the OpenPose_JSONs directory.

3. Edit Paths in json_copy.py:

   • Open json_copy.py in a text editor.
   • Update the following paths:

     path_to_json_files = 'Path/To/Your/OpenPose_JSONs/'  # Replace with the path to your OpenPose JSONs
     with open('Path/To/Your/Template/NFD_root-15.json') as f:  # Replace with the path to your COCO template
     with open('Path/To/Your/Remapped_COCO_Output/NFD_root-15.json', 'w') as f:  # Replace with output path

   • Save the file after updating the paths.

4. Run the json_copy.py Script:

   • Open a terminal or command prompt.
   • Navigate to the directory where json_copy.py is located:

     cd Path/To/Your/Scripts/

   • Run the script using Python:

     python json_copy.py

   • The script will read each JSON file from OpenPose_JSONs, remap the keypoints, and update the COCO JSON file in Remapped_COCO_Output.

5. (Optional) Clean Up Files Using file_handler.py:

   • If you need to clean up redundant files, update paths in file_handler.py to your file directories.
   • Run the script in a similar manner:

     python file_handler.py

   • This script will delete unnecessary image and JSON files based on the criteria specified.

Notes on Running the Scripts

• Script Execution Order: Always run json_copy.py first to perform the keypoint remapping and COCO annotation conversion.
• Backup Your Data: Before running file_handler.py, make sure to back up your data to avoid accidental deletions.

Verification

• Open the remapped COCO JSON file in Remapped_COCO_Output and inspect the annotations to ensure they follow the COCO keypoint format.

---