DataFlow-CV

🌊 Where Vibe Coding meets CV data. Convert, visualize & evaluate datasets — built with the flow of Claude Code.

A computer vision dataset processing library for seamless format conversion, visualization, and evaluation between YOLO, LabelMe, and COCO annotation formats. Designed for researchers and developers working with multi-format annotation pipelines.

graph LR
    A[YOLO<br/>.txt] -->|convert| D[DataFlow-CV]
    B[LabelMe<br/>.json] -->|convert| D
    C[COCO<br/>.json] -->|convert| D
    D -->|visualize| E[🎨 Rendered<br/>Images]
    D -->|evaluate| F[📊 mAP / AR<br/>Metrics]

Loading

---

✨ Features

🔄 Format Conversion	Convert between YOLO, LabelMe, and COCO in any direction — 6 conversion paths, plus prediction file support (outputs standard list-format COCO predictions)
🎯 Detection & Segmentation	Handle both object detection (bbox) and instance segmentation (polygon/RLE) annotations
🎨 Visualization	Render annotations with OpenCV — color-coded classes, semi-transparent masks, display & save modes
📊 Evaluation	COCO-standard 12-metric output (mAP, AP50, AP75, AR) via pycocotools, with per-class breakdowns
💻 Command-line Interface	Intuitive CLI with convert, visualize, and evaluate subcommands — positional args, rich --help
🐍 Python API	Programmatic access for integration into larger ML pipelines
📝 Verbose Logging	File-based debug logging with timestamps — toggle with --verbose
🖥️ Headless Mode	Server/Docker-friendly: --no-display + --save for off-screen rendering
🛡️ Flexible Error Handling	Strict mode (abort on error) or lenient mode (skip & continue with warnings) via --no-strict

---

📦 Installation

From PyPI

pip install dataflow-cv

From Source

git clone https://github.com/zjykzj/DataFlow-CV.git
cd DataFlow-CV

# Regular installation
pip install .

# Editable installation (for development)
pip install -e .

💡 Tip: When installed in editable mode, use python -m dataflow.cli instead of the dataflow-cv command.

Optional Dependencies

Dependency	Purpose	Install
pycocotools	COCO RLE segmentation + evaluation	pip install pycocotools

---

🚀 Quick Start

Command-line Interface

All required parameters (image directories, label directories, class files, output paths) are positional arguments for better usability. Use --help on any subcommand for detailed usage.

🔄 Format Conversion

# YOLO → COCO
dataflow-cv convert yolo2coco images/ yolo_labels/ classes.txt output.json

# YOLO → COCO (with RLE encoding)
dataflow-cv convert yolo2coco images/ yolo_labels/ classes.txt output.json --do-rle

# YOLO → LabelMe
dataflow-cv convert yolo2labelme images/ yolo_labels/ classes.txt labelme_json/

# LabelMe → YOLO
dataflow-cv convert labelme2yolo labelme_json/ classes.txt yolo_labels/

# LabelMe → COCO
dataflow-cv convert labelme2coco labelme_json/ classes.txt output.json

# COCO → YOLO
dataflow-cv convert coco2yolo input.json yolo_labels/

# COCO → LabelMe
dataflow-cv convert coco2labelme input.json labelme_json/

# YOLO predictions → COCO (output: plain JSON list — prediction format)
dataflow-cv convert yolo2coco --prediction images/ yolo_preds/ classes.txt pred.json

# Options
dataflow-cv convert yolo2coco --verbose images/ labels/ classes.txt output.json
dataflow-cv convert yolo2coco --no-strict images/ labels/ classes.txt output.json

🎨 Visualization

# Visualize YOLO annotations
dataflow-cv visualize yolo images/ yolo_labels/ classes.txt --save visualized/

# Visualize LabelMe annotations
dataflow-cv visualize labelme images/ labelme_json/ --save visualized/

# Visualize COCO annotations
dataflow-cv visualize coco images/ coco_annotations.json --save visualized/

# Verbose logging + headless mode
dataflow-cv visualize yolo --verbose --no-display images/ yolo_labels/ classes.txt --save visualized/

📊 Evaluation

Evaluate object detection and instance segmentation model outputs using COCO-standard metrics. Two COCO-format JSON files are required:

File	Role	Format	Source
anno.json	Ground Truth (GT) — reference annotations	Full COCO dict (images, annotations, categories)	yolo2coco (label mode)
pred.json	Detection (DT) — model predictions	Plain JSON list of annotation dicts (with score)	yolo2coco --prediction, Detectron2, MMDetection

① Preparing Evaluation Data

If your annotations and predictions are in YOLO format, convert them to COCO JSON first:

# Step 1: YOLO ground truth labels → COCO GT (anno.json)
#   Label format:   class_id cx cy w h               ← 5 tokens (detection)
#                   class_id x1 y1 ... xn yn          ← odd tokens (segmentation)
dataflow-cv convert yolo2coco images/ yolo_labels/ classes.txt anno.json

# Step 2: YOLO predictions → COCO DT (pred.json)
#   Prediction fmt: class_id cx cy w h confidence     ← 6 tokens (detection)
#                   class_id x1 y1 ... xn yn confidence ← even tokens (segmentation)
dataflow-cv convert yolo2coco --prediction images/ yolo_preds/ classes.txt pred.json

⚠️ Important: YOLO label files (GT) use odd token counts, while prediction files (DT) use even token counts with a trailing confidence. The --prediction flag is required for DT — it outputs a plain JSON list of annotation dicts (not a full COCO dict with images/categories). Mixed label/prediction files in the same directory are not supported.

ℹ️ Note: The --prediction flag is only available for yolo2coco. labelme2coco does not support prediction conversion — LabelMe files (.json) have no label vs prediction format distinction, so there is no equivalent prediction source format to convert from.

② Detection vs Segmentation — Format Requirements

Field	Detection GT	Detection DT	Segmentation GT	Segmentation DT
bbox	✅ Required	✅ Required	✅ Required (for area)	✅ Required (for area)
score	—	✅ Required	—	✅ Required
segmentation	❌ Not required	❌ Not required	✅ Required	✅ Required
area	⚪ Recommended	⚪ Recommended	✅ Required	✅ Required
iscrowd	⚪ Optional	—	⚪ Optional	—

• Object Detection (iouType='bbox'): Bounding box overlap evaluation. Only bbox + score mandatory in DT.
• Instance Segmentation (iouType='segm'): Mask overlap evaluation. GT and DT must include segmentation (polygon or RLE), area, and bbox.

③ CLI Commands

# Object detection evaluation (bbox IoU)
dataflow-cv evaluate detection anno.json pred.json

# Verbose per-class breakdown
dataflow-cv evaluate detection --verbose anno.json pred.json

# With P/R/F1 at IoU=0.5
dataflow-cv evaluate detection --prf1 --prf1-iou 0.5 anno.json pred.json

# Instance segmentation evaluation (mask IoU)
dataflow-cv evaluate segmentation anno.json pred.json

# Save results as JSON
dataflow-cv evaluate detection --output results.json anno.json pred.json

④ End-to-End Workflow

# Complete pipeline: YOLO → COCO → Evaluation
dataflow-cv convert yolo2coco images/ yolo_labels/ classes.txt anno.json
dataflow-cv convert yolo2coco --prediction images/ yolo_preds/ classes.txt pred.json
dataflow-cv evaluate detection --verbose --prf1 anno.json pred.json

🐍 Python API

from dataflow.convert import YoloAndCocoConverter
from dataflow.visualize import YOLOVisualizer
from dataflow.evaluate import DetectionEvaluator, compute_pr_f1

# ── Convert ──────────────────────────────────────────
# YOLO labels → COCO (label mode)
converter = YoloAndCocoConverter(source_to_target=True, verbose=True, strict_mode=True)
result = converter.convert(
    source_path="yolo_labels/", target_path="anno.json",
    class_file="classes.txt", image_dir="images/",
)

# YOLO predictions → COCO (prediction mode)
converter = YoloAndCocoConverter(source_to_target=True, prediction=True)
result = converter.convert(
    source_path="yolo_preds/", target_path="pred.json",
    class_file="classes.txt", image_dir="images/",
)

# ── Visualize ────────────────────────────────────────
visualizer = YOLOVisualizer(
    label_dir="yolo_labels/", image_dir="images/",
    class_file="classes.txt", is_show=True, is_save=True,
    output_dir="visualized/", verbose=True, strict_mode=True,
)
result = visualizer.visualize()

# ── Evaluate ─────────────────────────────────────────
evaluator = DetectionEvaluator(verbose=True)
result = evaluator.evaluate("anno.json", "pred.json")
print(f"AP: {result.metrics.ap:.3f}, AP50: {result.metrics.ap50:.3f}")

# Quick P/R/F1 at IoU=0.5
prf1 = compute_pr_f1("anno.json", "pred.json", iou_threshold=0.5)
print(f"F1: {prf1.overall.f1_score:.3f}")

📂 See the samples/ directory for complete examples: samples/visualize/ (YOLO, LabelMe, COCO demos), samples/convert/ (conversion examples).

---

📖 Documentation

Resource	Description
CLAUDE.md	Architecture overview, development guide, and known gotchas
CHANGELOG.md	Version history and breaking changes
specs/evaluate/	Evaluation metric contracts — IoU, matching, AP/mAP/AR
specs/formats/	External format contracts — YOLO, LabelMe, COCO, conversion rules
specs/modules/	Internal module architecture, interface contracts, dependency constraints

💡 Key Concepts

• Format-Native Coordinates: Coordinates stored in each format's native representation — YOLO normalized [0,1] center-based, LabelMe/COCO absolute pixels top-left. Check DatasetAnnotations.format to determine semantics.
• Explicit Coordinate Transforms: Converters handle all coordinate transformations between formats — no hidden normalization.
• Strict Mode: Validation errors raise exceptions by default. Disable with --no-strict (CLI) or strict_mode=False (API).
• Verbose Logging: Detailed debug logs saved to files when --verbose is used. The CLI prints the log file path after each operation.
• Headless Support: Use --no-display for servers/Docker; pair with --save to output visualization images without a window.
• Keyboard Shortcuts: During visualization — q/ESC to exit, Enter/Space to advance, any other key to continue.
• Color Management: Each class ID gets a unique color from an HSV-based palette (up to 1000 classes) for consistent visualization.
• Evaluation Metrics: COCO-standard 12-metric output with optional per-class breakdown and P/R/F1 computation.
• Prediction Files: YOLO prediction files use 6 tokens (detection) or even tokens (segmentation) vs 5/odd for labels. --prediction outputs a plain JSON list of annotation dicts — the standard prediction exchange format compatible with pycocotools loadRes().

---

🔧 Development

For detailed developer guidance including advanced test commands, debugging, and architecture overview, see CLAUDE.md.

🧪 Testing

370 tests, 75% code coverage (3912 statements).

pytest                                    # All tests
pytest --cov=dataflow --cov-report=term   # With coverage
pytest tests/convert/test_yolo_and_coco.py  # Single module
pytest tests/evaluate/test_evaluator.py     # Single module

📊 Coverage by module

Module	Coverage	Highlights
dataflow/label/	68%	models (87%), coco_handler (75%), labelme_handler (70%), yolo_handler (58%)
dataflow/convert/	87%	yolo_and_coco (90%), labelme_and_yolo (86%), coco_and_labelme (87%), rle (80%), base (83%), utils (92%)
dataflow/visualize/	81%	yolo_vis (100%), labelme_vis (100%), coco_vis (97%), base (74%)
dataflow/evaluate/	88%	evaluator (100%), metrics (96%), result (99%), base (91%), utils (69%)
dataflow/cli/	59%	main (96%), convert cmd (48%), evaluate cmd (24%), visualize cmd (84%), utils (86%)
dataflow/util/	93%	logging (98%), file_util (84%)

🎨 Code Quality

pip install -e .[dev]        # Install dev dependencies
black dataflow tests samples  # Format
isort dataflow tests samples  # Sort imports
mypy dataflow                 # Type check
flake8 dataflow tests samples # Lint

🔗 Pre-commit Hooks (Optional)

pip install pre-commit
pre-commit install            # Install git hooks (run once)

# After this, every `git commit` auto-runs:
#   black → isort → flake8 → whitespace checks

pre-commit run --all-files    # Manual run against all files

📁 Project Structure

dataflow/
├── label/           # Annotation handlers + data models
├── convert/         # Format converters + RLE utility
├── visualize/       # OpenCV-based rendering
├── evaluate/        # pycocotools-based metrics
├── util/            # Logging & file utilities
└── cli/             # CLI entry point, commands, validation
tests/               # Unit & integration tests
samples/             # Python API usage examples
assets/              # Test data (det/seg by format)
specs/               # Canonical specifications (evaluate/ + formats/ + modules/)

---

🤝 Contributing

Contributions are welcome! Please review CLAUDE.md for architecture and development patterns before contributing.

1. 🍴 Fork the repository
2. 🌿 Create a feature branch
3. ✏️ Make your changes
4. 🧪 Add or update tests as needed
5. ✅ Ensure code passes formatting and linting checks
6. 📬 Submit a pull request

---

📄 License

This project is licensed under the MIT License — see LICENSE for details.

---

🙏 Acknowledgments

• Thanks to the creators of YOLO, LabelMe, and COCO formats for establishing these annotation standards
• Built with OpenCV, NumPy, Click, and pycocotools
• Inspired by the need for seamless format conversion in multi-tool CV pipelines