DOWNLOAD_INSTRUCTIONS.md

Download COCO Dataset to Local Computer

Archive Created ✅

File: /workspace/soccer_coach_cv_coco_dataset.tar.gz
Size: ~5.0 GB (compressed from 5.5 GB)
Contains: Complete COCO format dataset (train + val splits)

Download Methods

Method 1: SCP (Command Line)

From your local terminal, run:

scp user@your-server:/workspace/soccer_coach_cv_coco_dataset.tar.gz ./

Replace:

• user@your-server with your actual SSH connection details
• If using RunPod, it might be: root@your-pod-ip

Method 2: SFTP

sftp user@your-server
get /workspace/soccer_coach_cv_coco_dataset.tar.gz
exit

Method 3: Web Interface (RunPod/Cloud IDE)

1. Open the file browser in your web interface
2. Navigate to /workspace/
3. Find soccer_coach_cv_coco_dataset.tar.gz
4. Right-click → Download

Method 4: VS Code Remote

If using VS Code with Remote SSH:

1. Open VS Code Remote Explorer
2. Navigate to /workspace/soccer_coach_cv_coco_dataset.tar.gz
3. Right-click → Download

Extract on Local Machine

After downloading, extract the archive:

tar -xzf soccer_coach_cv_coco_dataset.tar.gz

This will create:

soccer_coach_cv_coco_dataset/
├── train/
│   ├── images/          (1,977 images)
│   └── annotations/
│       └── annotations.json
└── val/
    ├── images/          (1,012 images)
    └── annotations/
        └── annotations.json

Dataset Contents

• Training: 1,977 images, 10,973 annotations
• Validation: 1,012 images, 5,596 annotations
• Classes: player (class 0), ball (class 1)
• Format: COCO JSON format
• Image size: 1920x1080

Verification

After extraction, verify the dataset:

ls soccer_coach_cv_coco_dataset/train/images/ | wc -l  # Should show 1977
ls soccer_coach_cv_coco_dataset/val/images/ | wc -l    # Should show 1012