dataflow-cv 2.0.1__tar.gz → 3.0.0__tar.gz

{dataflow_cv-2.0.1/dataflow_cv.egg-info → dataflow_cv-3.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dataflow-cv
-Version: 2.0.1
+Version: 3.0.0
 Summary: A computer vision dataset processing library — analyse, convert, visualize, and evaluate annotations across YOLO, LabelMe, and COCO formats
 Author: DataFlow-CV Team
 License: MIT
@@ -156,6 +156,12 @@ dataflow-cv convert yolo2coco --verbose images/ labels/ classes.txt output.json
 dataflow-cv convert yolo2coco --no-strict images/ labels/ classes.txt output.json
 ```
 
+> 📂 **COCO→YOLO / COCO→LabelMe output naming**: each `.txt` / `.json` is named by the
+> image file stem — leading zeros preserved (`000001.jpg` → `000001.txt`). The
+> converter generates `labels/` (or the `.json` files) + `classes.txt`; the
+> `images/` directory is created but left **empty** — images are never copied,
+> place your image files there yourself.
+
 #### 🎨 Visualization
 
 ```bash
@@ -231,7 +237,13 @@ Two evaluation modes, distinguished by how overlap is measured:
 
 ```python
 from dataflow.util.logging import LogConfig
-from dataflow.analyse import StatsAnalyser, SplitAnalyser, FilterAnalyser, PartitionAnalyser, SampleAnalyser
+from dataflow.analyse import (
+    StatsAnalyser,
+    SplitAnalyser,
+    FilterAnalyser,
+    PartitionAnalyser,
+    SampleAnalyser,
+)
 from dataflow.convert import YoloAndCocoConverter
 from dataflow.visualize import YOLOVisualizer
 from dataflow.evaluate import DetectionEvaluator, compute_pr_f1
@@ -247,37 +259,50 @@ print(f"{result.data.total_files} images, {result.data.total_annotations} object
 # Train/test split (YOLO / LabelMe)
 splitter = SplitAnalyser(log_config=log_cfg)
 result = splitter.analyse(
-    output_dir="output/", ratio=0.8, seed=42,
-    label_dir="yolo_labels/", class_file="classes.txt",
+    output_dir="output/",
+    ratio=0.8,
+    seed=42,
+    label_dir="yolo_labels/",
+    class_file="classes.txt",
 )
 print(f"Train: {result.data.train_count}, Val: {result.data.val_count}")
 
 # Split with images (both mode — labels drive, images follow by stem)
 result = splitter.analyse(
-    output_dir="output/", ratio=0.8, seed=42,
-    label_dir="yolo_labels/", image_dir="images/",
+    output_dir="output/",
+    ratio=0.8,
+    seed=42,
+    label_dir="yolo_labels/",
+    image_dir="images/",
     class_file="classes.txt",
 )
 
 # Category filter (keep / remap categories per new classes.txt)
 filterer = FilterAnalyser(log_config=log_cfg)
 result = filterer.analyse(
-    "yolo_labels/", original_class_file="classes.txt",
-    new_class_file="classes_new.txt", output_dir="filtered/",
+    "yolo_labels/",
+    original_class_file="classes.txt",
+    new_class_file="classes_new.txt",
+    output_dir="filtered/",
 )
 
 # N-way partition (YOLO / LabelMe labels; images follow by stem)
 partitioner = PartitionAnalyser(log_config=log_cfg)
 result = partitioner.analyse(
-    output_dir="parts/", num=4,
-    label_dir="yolo_labels/", image_dir="images/",
+    output_dir="parts/",
+    num=4,
+    label_dir="yolo_labels/",
+    image_dir="images/",
 )
 
 # File sampling (labels, images, or both — random or sequential)
 sampler = SampleAnalyser(log_config=log_cfg)
 result = sampler.analyse(
-    output_dir="sampled/", count=10,
-    label_dir="yolo_labels/", shuffle=True, seed=42,
+    output_dir="sampled/",
+    count=10,
+    label_dir="yolo_labels/",
+    shuffle=True,
+    seed=42,
 )
 
 # ── Convert ──────────────────────────────────────────
@@ -285,22 +310,30 @@ result = sampler.analyse(
 log_cfg = LogConfig(name="convert", verbose=True)
 converter = YoloAndCocoConverter(source_to_target=True, log_config=log_cfg, strict_mode=True)
 result = converter.convert(
-    source_path="yolo_labels/", target_path="anno.json",
-    class_file="classes.txt", image_dir="images/",
+    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/",
+    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/", log_config=log_cfg,
+    label_dir="yolo_labels/",
+    image_dir="images/",
+    class_file="classes.txt",
+    is_show=True,
+    is_save=True,
+    output_dir="visualized/",
+    log_config=log_cfg,
 )
 result = visualizer.visualize()
 
@@ -370,7 +403,7 @@ For detailed developer guidance including advanced test commands, debugging, and
 
 ### 🧪 Testing
 
-**561 tests, 80% code coverage (5462 statements).**
+**606 tests, 80% code coverage (5532 statements).**
 
 ```bash
 pytest                                    # All tests
@@ -384,11 +417,11 @@ pytest tests/evaluate/test_evaluator.py     # Single module
 
 | Module | Coverage | Highlights |
 |--------|:--------:|------------|
-| `dataflow/label/` | 71% | models (84%), base (82%), utils (78%), coco_handler (74%), labelme_handler (71%), yolo_handler (61%) |
-| `dataflow/analyse/` | 84% | base (99%), log_templates (92%), sample (87%), utils (85%), split (85%), stats (83%), filter (76%), partition (74%) |
-| `dataflow/convert/` | 85% | labelme_and_yolo (93%), yolo_and_coco (89%), utils (87%), coco_and_labelme (86%), base (80%), rle (80%) |
+| `dataflow/label/` | 71% | models (84%), base (82%), utils (83%), coco_handler (74%), labelme_handler (71%), yolo_handler (61%) |
+| `dataflow/analyse/` | 84% | base (99%), log_templates (92%), sample (87%), utils (84%), split (85%), stats (82%), filter (76%), partition (74%) |
+| `dataflow/convert/` | 86% | labelme_and_yolo (93%), yolo_and_coco (89%), utils (88%), coco_and_labelme (88%), log_templates (85%), base (81%), rle (80%) |
 | `dataflow/visualize/` | 81% | yolo_vis (100%), labelme_vis (100%), coco_vis (93%), base (76%) |
-| `dataflow/evaluate/` | 87% | evaluator (100%), result (99%), metrics (93%), base (90%), utils (67%) |
+| `dataflow/evaluate/` | 90% | evaluator (100%), result (99%), metrics (93%), base (90%), utils (87%) |
 | `dataflow/cli/` | 74% | main (96%), visualize cmd (87%), utils (87%), evaluate cmd (83%), analyse cmd (65%), convert cmd (52%) |
 | `dataflow/util/` | 100% | logging (100%) |
 
@@ -397,11 +430,10 @@ pytest tests/evaluate/test_evaluator.py     # Single module
 ### 🎨 Code Quality
 
 ```bash
-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
+pip install -e .[dev]                       # Install dev dependencies
+ruff check dataflow tests samples           # Lint
+ruff format --check dataflow tests samples  # Format check
+mypy dataflow                               # Type check
 ```
 
 ### 🔗 Pre-commit Hooks (Optional)
@@ -411,7 +443,7 @@ pip install pre-commit
 pre-commit install            # Install git hooks (run once)
 
 # After this, every `git commit` auto-runs:
-#   black → isort → flake8 → whitespace checks
+#   ruff (lint, auto-fix) → ruff format → whitespace checks
 
 pre-commit run --all-files    # Manual run against all files
 ```
@@ -427,7 +459,7 @@ dataflow/
 ├── evaluate/        # pycocotools-based metrics, log templates
 ├── util/            # Unified logging (LogManager + format helpers)
 └── cli/             # CLI entry point, commands, validation
-tests/               # Unit & integration tests (561 tests, conftest fixtures)
+tests/               # Unit & integration tests (606 tests, conftest fixtures)
 samples/             # Python API usage examples (analyse, convert, visualize, evaluate, cli)
 assets/              # Test data (det/seg by format)
 specs/               # Canonical specifications (evaluate/ + formats/ + modules/)

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/README.md

@@ -117,6 +117,12 @@ dataflow-cv convert yolo2coco --verbose images/ labels/ classes.txt output.json
 dataflow-cv convert yolo2coco --no-strict images/ labels/ classes.txt output.json
 ```
 
+> 📂 **COCO→YOLO / COCO→LabelMe output naming**: each `.txt` / `.json` is named by the
+> image file stem — leading zeros preserved (`000001.jpg` → `000001.txt`). The
+> converter generates `labels/` (or the `.json` files) + `classes.txt`; the
+> `images/` directory is created but left **empty** — images are never copied,
+> place your image files there yourself.
+
 #### 🎨 Visualization
 
 ```bash
@@ -192,7 +198,13 @@ Two evaluation modes, distinguished by how overlap is measured:
 
 ```python
 from dataflow.util.logging import LogConfig
-from dataflow.analyse import StatsAnalyser, SplitAnalyser, FilterAnalyser, PartitionAnalyser, SampleAnalyser
+from dataflow.analyse import (
+    StatsAnalyser,
+    SplitAnalyser,
+    FilterAnalyser,
+    PartitionAnalyser,
+    SampleAnalyser,
+)
 from dataflow.convert import YoloAndCocoConverter
 from dataflow.visualize import YOLOVisualizer
 from dataflow.evaluate import DetectionEvaluator, compute_pr_f1
@@ -208,37 +220,50 @@ print(f"{result.data.total_files} images, {result.data.total_annotations} object
 # Train/test split (YOLO / LabelMe)
 splitter = SplitAnalyser(log_config=log_cfg)
 result = splitter.analyse(
-    output_dir="output/", ratio=0.8, seed=42,
-    label_dir="yolo_labels/", class_file="classes.txt",
+    output_dir="output/",
+    ratio=0.8,
+    seed=42,
+    label_dir="yolo_labels/",
+    class_file="classes.txt",
 )
 print(f"Train: {result.data.train_count}, Val: {result.data.val_count}")
 
 # Split with images (both mode — labels drive, images follow by stem)
 result = splitter.analyse(
-    output_dir="output/", ratio=0.8, seed=42,
-    label_dir="yolo_labels/", image_dir="images/",
+    output_dir="output/",
+    ratio=0.8,
+    seed=42,
+    label_dir="yolo_labels/",
+    image_dir="images/",
     class_file="classes.txt",
 )
 
 # Category filter (keep / remap categories per new classes.txt)
 filterer = FilterAnalyser(log_config=log_cfg)
 result = filterer.analyse(
-    "yolo_labels/", original_class_file="classes.txt",
-    new_class_file="classes_new.txt", output_dir="filtered/",
+    "yolo_labels/",
+    original_class_file="classes.txt",
+    new_class_file="classes_new.txt",
+    output_dir="filtered/",
 )
 
 # N-way partition (YOLO / LabelMe labels; images follow by stem)
 partitioner = PartitionAnalyser(log_config=log_cfg)
 result = partitioner.analyse(
-    output_dir="parts/", num=4,
-    label_dir="yolo_labels/", image_dir="images/",
+    output_dir="parts/",
+    num=4,
+    label_dir="yolo_labels/",
+    image_dir="images/",
 )
 
 # File sampling (labels, images, or both — random or sequential)
 sampler = SampleAnalyser(log_config=log_cfg)
 result = sampler.analyse(
-    output_dir="sampled/", count=10,
-    label_dir="yolo_labels/", shuffle=True, seed=42,
+    output_dir="sampled/",
+    count=10,
+    label_dir="yolo_labels/",
+    shuffle=True,
+    seed=42,
 )
 
 # ── Convert ──────────────────────────────────────────
@@ -246,22 +271,30 @@ result = sampler.analyse(
 log_cfg = LogConfig(name="convert", verbose=True)
 converter = YoloAndCocoConverter(source_to_target=True, log_config=log_cfg, strict_mode=True)
 result = converter.convert(
-    source_path="yolo_labels/", target_path="anno.json",
-    class_file="classes.txt", image_dir="images/",
+    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/",
+    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/", log_config=log_cfg,
+    label_dir="yolo_labels/",
+    image_dir="images/",
+    class_file="classes.txt",
+    is_show=True,
+    is_save=True,
+    output_dir="visualized/",
+    log_config=log_cfg,
 )
 result = visualizer.visualize()
 
@@ -331,7 +364,7 @@ For detailed developer guidance including advanced test commands, debugging, and
 
 ### 🧪 Testing
 
-**561 tests, 80% code coverage (5462 statements).**
+**606 tests, 80% code coverage (5532 statements).**
 
 ```bash
 pytest                                    # All tests
@@ -345,11 +378,11 @@ pytest tests/evaluate/test_evaluator.py     # Single module
 
 | Module | Coverage | Highlights |
 |--------|:--------:|------------|
-| `dataflow/label/` | 71% | models (84%), base (82%), utils (78%), coco_handler (74%), labelme_handler (71%), yolo_handler (61%) |
-| `dataflow/analyse/` | 84% | base (99%), log_templates (92%), sample (87%), utils (85%), split (85%), stats (83%), filter (76%), partition (74%) |
-| `dataflow/convert/` | 85% | labelme_and_yolo (93%), yolo_and_coco (89%), utils (87%), coco_and_labelme (86%), base (80%), rle (80%) |
+| `dataflow/label/` | 71% | models (84%), base (82%), utils (83%), coco_handler (74%), labelme_handler (71%), yolo_handler (61%) |
+| `dataflow/analyse/` | 84% | base (99%), log_templates (92%), sample (87%), utils (84%), split (85%), stats (82%), filter (76%), partition (74%) |
+| `dataflow/convert/` | 86% | labelme_and_yolo (93%), yolo_and_coco (89%), utils (88%), coco_and_labelme (88%), log_templates (85%), base (81%), rle (80%) |
 | `dataflow/visualize/` | 81% | yolo_vis (100%), labelme_vis (100%), coco_vis (93%), base (76%) |
-| `dataflow/evaluate/` | 87% | evaluator (100%), result (99%), metrics (93%), base (90%), utils (67%) |
+| `dataflow/evaluate/` | 90% | evaluator (100%), result (99%), metrics (93%), base (90%), utils (87%) |
 | `dataflow/cli/` | 74% | main (96%), visualize cmd (87%), utils (87%), evaluate cmd (83%), analyse cmd (65%), convert cmd (52%) |
 | `dataflow/util/` | 100% | logging (100%) |
 
@@ -358,11 +391,10 @@ pytest tests/evaluate/test_evaluator.py     # Single module
 ### 🎨 Code Quality
 
 ```bash
-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
+pip install -e .[dev]                       # Install dev dependencies
+ruff check dataflow tests samples           # Lint
+ruff format --check dataflow tests samples  # Format check
+mypy dataflow                               # Type check
 ```
 
 ### 🔗 Pre-commit Hooks (Optional)
@@ -372,7 +404,7 @@ pip install pre-commit
 pre-commit install            # Install git hooks (run once)
 
 # After this, every `git commit` auto-runs:
-#   black → isort → flake8 → whitespace checks
+#   ruff (lint, auto-fix) → ruff format → whitespace checks
 
 pre-commit run --all-files    # Manual run against all files
 ```
@@ -388,7 +420,7 @@ dataflow/
 ├── evaluate/        # pycocotools-based metrics, log templates
 ├── util/            # Unified logging (LogManager + format helpers)
 └── cli/             # CLI entry point, commands, validation
-tests/               # Unit & integration tests (561 tests, conftest fixtures)
+tests/               # Unit & integration tests (606 tests, conftest fixtures)
 samples/             # Python API usage examples (analyse, convert, visualize, evaluate, cli)
 assets/              # Test data (det/seg by format)
 specs/               # Canonical specifications (evaluate/ + formats/ + modules/)

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/__init__.py

@@ -1,4 +1,4 @@
 """DataFlow-CV: A computer vision dataset processing library."""
 
-__version__ = "2.0.1"
+__version__ = "3.0.0"
 __author__ = "DataFlow-CV Team"

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/convert/base.py

@@ -8,7 +8,7 @@ import datetime
 from abc import ABC, abstractmethod
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import Any, Dict, List, Optional
+from typing import Any, Dict, List, Optional, Set
 
 from ..label.base import AnnotationResult, BaseAnnotationHandler
 from ..label.models import AnnotationFormat, DatasetAnnotations, ImageAnnotation
@@ -215,6 +215,7 @@ class BaseConverter(ABC):
 
         num_images = 0
         num_objects = 0
+        written_ids: Set[str] = set()  # Detect duplicate output stems
 
         try:
             # 1. Validate inputs
@@ -238,6 +239,38 @@ class BaseConverter(ABC):
 
             for image_ann in source_handler.iter_images():
                 target_ann = self._convert_single_image(image_ann, **kwargs)
+
+                # Output-name notices: COCO-source converters rewrite
+                # image_id to the file_name stem (e.g. "1" → "000001"), and
+                # file_name directory components are dropped (per-file
+                # outputs are flat per the YOLO/LabelMe format specs).
+                renamed = target_ann.image_id != image_ann.image_id
+                flattened = self.source_format == "coco" and (
+                    Path(str(image_ann.image_path).replace("\\", "/")).parent != Path(".")
+                )
+                if renamed or flattened:
+                    parts = []
+                    if renamed:
+                        parts.append(f"renamed {image_ann.image_id} → {target_ann.image_id}")
+                    if flattened:
+                        parts.append(f"subdirectory flattened from '{image_ann.image_path}'")
+                    notice = f"Output '{target_ann.image_id}': " + ", ".join(parts)
+                    result.add_warning(notice)
+                    self._log_warning(notice)
+
+                # Duplicate output stem detection: two source images sharing
+                # a basename (e.g. in different subdirectories) would silently
+                # overwrite each other's output file.
+                if target_ann.image_id in written_ids:
+                    dup = (
+                        f"Duplicate output stem '{target_ann.image_id}': "
+                        f"{image_ann.image_path} overwrites a previously written file"
+                    )
+                    result.add_warning(dup)
+                    self._log_warning(dup)
+                else:
+                    written_ids.add(target_ann.image_id)
+
                 write_result = target_handler.write_one(target_ann, write_dir)
                 if not write_result.success:
                     err = f"Failed to write {target_ann.image_id}: {write_result.message}"

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/convert/coco_and_labelme.py

@@ -220,6 +220,16 @@ class CocoAndLabelMeConverter(BaseConverter):
         Only the structural representation differs — coordinate values pass
         through unchanged.
         """
+        # COCO → LabelMe only: rewrite image_id to the COCO file_name stem so
+        # the output .json shares the stem with the image file (LabelMe format
+        # requirement; leading zeros preserved). LabelMe → COCO must NOT
+        # rewrite — the COCO output id derives from the numeric image_id.
+        new_image_id = image_ann.image_id
+        if self.source_format == "coco":
+            from .utils import coco_file_name_to_image_id
+
+            new_image_id = coco_file_name_to_image_id(image_ann.image_path, image_ann.image_id)
+
         new_objects = []
         for obj in image_ann.objects:
             new_bbox = None
@@ -251,7 +261,7 @@ class CocoAndLabelMeConverter(BaseConverter):
             )
 
         return ImageAnnotation(
-            image_id=image_ann.image_id,
+            image_id=new_image_id,
             image_path=image_ann.image_path,
             width=image_ann.width,
             height=image_ann.height,

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/convert/log_templates.py

@@ -65,9 +65,62 @@ def format_convert_phase(phase: str, stats: Dict[str, Any]) -> str:
     return "\n".join(lines)
 
 
+def format_output_layout(target_path: str) -> str:
+    """Return a compact summary of the generated output layout.
+
+    Scans the actual target path and lists its top-level entries:
+    directories show their recursive file count (``labels/  4 files``);
+    empty directories show ``empty`` — an empty ``images/`` additionally
+    notes that images are not copied; single-file targets (COCO JSON)
+    show the file name.
+
+    Args:
+        target_path: Target output path of a conversion.
+
+    Returns:
+        ``"Output layout:"`` block string, or ``""`` when the target
+        path does not exist.
+    """
+    from pathlib import Path
+
+    target = Path(target_path)
+    if not target.exists():
+        return ""
+
+    lines = ["Output layout:"]
+    if target.is_file():
+        lines.append(f"  {target.name}  (single file)")
+        return "\n".join(lines)
+
+    try:
+        entries = sorted(target.iterdir(), key=lambda p: (p.is_file(), p.name.lower()))
+    except OSError:
+        return ""
+
+    for entry in entries:
+        if entry.is_dir():
+            count = sum(1 for p in entry.rglob("*") if p.is_file())
+            if count == 0:
+                if entry.name == "images":
+                    lines.append(
+                        f"  {entry.name}/  empty — images are not copied; "
+                        "place your image files here"
+                    )
+                else:
+                    lines.append(f"  {entry.name}/  empty")
+            else:
+                lines.append(f"  {entry.name}/  {count} files")
+        else:
+            lines.append(f"  {entry.name}")
+    return "\n".join(lines)
+
+
 def format_convert_result(result: Any) -> str:
     """Return a final result block for a completed conversion.
 
+    The block is followed by an output layout summary (see
+    ``format_output_layout()``) when the target path exists.
+
     Args:
         result: A ``ConversionResult`` instance.
 
@@ -90,4 +143,11 @@ def format_convert_result(result: Any) -> str:
     if result.warnings:
         items["Warnings"] = len(result.warnings)
 
-    return format_result_block(status, items, log_path=result.log_path)
+    block = format_result_block(status, items, log_path=result.log_path)
+
+    if getattr(result, "target_path", None):
+        layout = format_output_layout(result.target_path)
+        if layout:
+            block = f"{block}\n\n{layout}"
+
+    return block

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/convert/utils.py

@@ -130,6 +130,42 @@ def absolute_pixel_to_yolo(
 # ---------------------------------------------------------------------------
 
 
+def coco_file_name_to_image_id(file_name: Optional[str], fallback_id: str) -> str:
+    """Derive the per-file output ``image_id`` from a COCO ``file_name``.
+
+    COCO ``file_name`` is the authoritative source of the image file stem
+    (``id`` is only a numeric reference key and carries no filename
+    information).  Per-file output formats (YOLO `.txt`, LabelMe `.json`)
+    must share the stem with the image file (see `spec_yolo_format.md` /
+    `spec_labelme_format.md`), so COCO-source converters rewrite ``image_id``
+    to this stem before ``write_one()``.
+
+    Rules:
+    - Backslashes are normalized to ``/`` (Windows-style paths)
+    - Subdirectory components are dropped (output is flat; the
+      `spec_label.md` image_id invariant forbids path separators)
+    - Leading zeros are preserved (``000001.jpg`` → ``000001``)
+    - Degenerate stems (``""``, ``"."``, ``".."``) fall back to *fallback_id*
+      (the numeric COCO id string)
+
+    Args:
+        file_name: COCO ``images[].file_name`` value. ``None`` is tolerated
+            defensively and treated as empty (falls back to *fallback_id*).
+        fallback_id: ``ImageAnnotation.image_id`` from the COCO handler
+            (numeric id string), used when the stem is degenerate.
+
+    Returns:
+        Basename stem of *file_name*, or *fallback_id*.
+    """
+    if not file_name:
+        return fallback_id
+    normalized = str(file_name).replace("\\", "/")
+    stem = Path(normalized).stem
+    if stem in ("", ".", ".."):
+        return fallback_id
+    return stem
+
+
 def ensure_coco_categories_for_streaming(
     converter: Any,
     source_handler: Any,

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/convert/yolo_and_coco.py

@@ -288,7 +288,7 @@ class YoloAndCocoConverter(BaseConverter):
 
     def _coco_to_yolo_one(self, img: ImageAnnotation) -> ImageAnnotation:
         """Convert single image: COCO absolute px → YOLO normalized center."""
-        from .utils import absolute_pixel_to_yolo
+        from .utils import absolute_pixel_to_yolo, coco_file_name_to_image_id
 
         new_objects = []
         for obj in img.objects:
@@ -306,8 +306,11 @@ class YoloAndCocoConverter(BaseConverter):
                 )
             )
 
+        # YOLO format requires label files to share the stem with the image
+        # file — derive the output name from the COCO file_name stem (which
+        # preserves leading zeros), not from the numeric COCO id.
         return ImageAnnotation(
-            image_id=img.image_id,
+            image_id=coco_file_name_to_image_id(img.image_path, img.image_id),
             image_path=img.image_path,
             width=img.width,
             height=img.height,

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/evaluate/utils.py

@@ -203,11 +203,17 @@ def _dataset_to_coco_dict(dataset: Any) -> Dict[str, Any]:
     # Preserve top-level keys from dataset_info
     info = dataset.dataset_info.copy()
 
-    # Reconstruct images
+    # Reconstruct images — image ids resolve with the same rule as the COCO
+    # handler write() (digit image_id → int, non-digit/colliding → dedicated
+    # counter, unique ids guaranteed — resolve_coco_image_ids()).
+    from dataflow.label.utils import resolve_coco_image_ids
+
+    id_map = resolve_coco_image_ids(img.image_id for img in dataset.images)
+
     images = []
     for img in dataset.images:
         img_entry = {
-            "id": int(img.image_id) if img.image_id.isdigit() else img.image_id,
+            "id": id_map[img.image_id],
             "file_name": img.image_path,
             "width": img.width,
             "height": img.height,
@@ -223,7 +229,7 @@ def _dataset_to_coco_dict(dataset: Any) -> Dict[str, Any]:
     annotations = []
     ann_id = 1
     for img in dataset.images:
-        image_id = int(img.image_id) if str(img.image_id).isdigit() else img.image_id
+        image_id = id_map[img.image_id]
         for obj in img.objects:
             ann_entry: Dict[str, Any] = {
                 "id": ann_id,

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/label/coco_handler.py

@@ -607,12 +607,18 @@ class CocoAnnotationHandler(BaseAnnotationHandler):
         images = []
         coco_annotations = []
         ann_id = 1
-        img_id_counter = 1
+
+        # Resolve format-native image_id strings to unique positive-int COCO
+        # ids (digit strings keep their value; non-digit and colliding ids
+        # get dedicated counter ids — see resolve_coco_image_ids()).
+        from .utils import resolve_coco_image_ids
+
+        id_map = resolve_coco_image_ids(img.image_id for img in annotations.images)
 
         for img in annotations.images:
             images.append(
                 {
-                    "id": int(img.image_id) if img.image_id.isdigit() else img_id_counter,
+                    "id": id_map[img.image_id],
                     "width": img.width,
                     "height": img.height,
                     "file_name": img.image_path,
@@ -625,7 +631,7 @@ class CocoAnnotationHandler(BaseAnnotationHandler):
 
             # Add object annotations
             for obj in img.objects:
-                coco_ann = self._object_to_coco_annotation(obj, img, ann_id, img_id_counter)
+                coco_ann = self._object_to_coco_annotation(obj, img, ann_id, id_map[img.image_id])
                 if coco_ann:
                     coco_annotations.append(coco_ann)
                     ann_id += 1
@@ -639,8 +645,6 @@ class CocoAnnotationHandler(BaseAnnotationHandler):
                         f"Skipping object {obj.class_name}: conversion to COCO format failed"
                     )
 
-            img_id_counter += 1
-
         # Prediction mode: output plain list of annotation dicts
         if self.prediction:
             return coco_annotations
@@ -661,9 +665,14 @@ class CocoAnnotationHandler(BaseAnnotationHandler):
         return result
 
     def _object_to_coco_annotation(
-        self, obj: ObjectAnnotation, img: ImageAnnotation, ann_id: int, img_id: int
+        self, obj: ObjectAnnotation, img: ImageAnnotation, ann_id: int, resolved_img_id: int
     ) -> Optional[Dict]:
-        """Convert ObjectAnnotation to COCO annotation dict."""
+        """Convert ObjectAnnotation to COCO annotation dict.
+
+        Args:
+            resolved_img_id: COCO image id resolved from ``img.image_id`` by
+                ``resolve_coco_image_ids()`` (see ``_prepare_coco_data``).
+        """
         try:
             # Determine segmentation format
             segmentation = None
@@ -767,10 +776,8 @@ class CocoAnnotationHandler(BaseAnnotationHandler):
                     bbox = [float(min_x), float(min_y), float(w), float(h)]
                     area = float(w * h)
 
-            image_id_val = int(img.image_id) if img.image_id.isdigit() else img_id
-
             ann_dict = {
-                "image_id": image_id_val,
+                "image_id": resolved_img_id,
                 "category_id": obj.class_id,
                 "segmentation": segmentation,
                 "area": area,

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/dataflow/label/utils.py

@@ -4,7 +4,52 @@ Utility functions for the label module.
 
 import hashlib
 from pathlib import Path
-from typing import Dict, Optional
+from typing import Dict, Iterable, Optional, Set
+
+
+def resolve_coco_image_ids(image_ids: Iterable[str]) -> Dict[str, int]:
+    """Resolve format-native ``image_id`` strings to unique positive-int COCO ids.
+
+    COCO ``images[].id`` must be an int and unique per image
+    (`spec_coco_format.md`).  ``ImageAnnotation.image_id`` is a string and may
+    be a file stem, so COCO writers resolve it:
+
+    - Digit-string ``image_id`` → its ``int`` value (preserves roundtrip ids)
+    - Non-digit ``image_id`` → a dedicated counter id (not the annotation
+      counter); counters skip values reserved by digit-string ids
+    - Uniqueness guaranteed: if a resolved id collides (e.g. ``"01"`` vs
+      ``"1"``), the later image gets a fresh counter id
+
+    Canonical implementation — used by both
+    ``CocoAnnotationHandler._prepare_coco_data()`` and
+    ``evaluate/utils._dataset_to_coco_dict()``.
+
+    Args:
+        image_ids: Iterable of image_id strings in output order.
+
+    Returns:
+        Mapping ``{image_id_string: resolved_int_id}``.
+    """
+    ids = [str(s) for s in image_ids]
+    # Values reserved by digit-string ids — counters must never take them,
+    # otherwise a later digit id would collide with an earlier counter.
+    digit_values: Set[int] = {int(s) for s in ids if s.isdigit()}
+    used: Set[int] = set()
+    mapping: Dict[str, int] = {}
+    counter = 1
+
+    for s in ids:
+        if s.isdigit() and int(s) not in used:
+            value = int(s)
+        else:
+            while counter in used or counter in digit_values:
+                counter += 1
+            value = counter
+            counter += 1
+        used.add(value)
+        mapping[s] = value
+
+    return mapping
 
 
 def parse_yolo_class_id(token: str) -> Optional[int]:

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0/dataflow_cv.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dataflow-cv
-Version: 2.0.1
+Version: 3.0.0
 Summary: A computer vision dataset processing library — analyse, convert, visualize, and evaluate annotations across YOLO, LabelMe, and COCO formats
 Author: DataFlow-CV Team
 License: MIT
@@ -156,6 +156,12 @@ dataflow-cv convert yolo2coco --verbose images/ labels/ classes.txt output.json
 dataflow-cv convert yolo2coco --no-strict images/ labels/ classes.txt output.json
 ```
 
+> 📂 **COCO→YOLO / COCO→LabelMe output naming**: each `.txt` / `.json` is named by the
+> image file stem — leading zeros preserved (`000001.jpg` → `000001.txt`). The
+> converter generates `labels/` (or the `.json` files) + `classes.txt`; the
+> `images/` directory is created but left **empty** — images are never copied,
+> place your image files there yourself.
+
 #### 🎨 Visualization
 
 ```bash
@@ -231,7 +237,13 @@ Two evaluation modes, distinguished by how overlap is measured:
 
 ```python
 from dataflow.util.logging import LogConfig
-from dataflow.analyse import StatsAnalyser, SplitAnalyser, FilterAnalyser, PartitionAnalyser, SampleAnalyser
+from dataflow.analyse import (
+    StatsAnalyser,
+    SplitAnalyser,
+    FilterAnalyser,
+    PartitionAnalyser,
+    SampleAnalyser,
+)
 from dataflow.convert import YoloAndCocoConverter
 from dataflow.visualize import YOLOVisualizer
 from dataflow.evaluate import DetectionEvaluator, compute_pr_f1
@@ -247,37 +259,50 @@ print(f"{result.data.total_files} images, {result.data.total_annotations} object
 # Train/test split (YOLO / LabelMe)
 splitter = SplitAnalyser(log_config=log_cfg)
 result = splitter.analyse(
-    output_dir="output/", ratio=0.8, seed=42,
-    label_dir="yolo_labels/", class_file="classes.txt",
+    output_dir="output/",
+    ratio=0.8,
+    seed=42,
+    label_dir="yolo_labels/",
+    class_file="classes.txt",
 )
 print(f"Train: {result.data.train_count}, Val: {result.data.val_count}")
 
 # Split with images (both mode — labels drive, images follow by stem)
 result = splitter.analyse(
-    output_dir="output/", ratio=0.8, seed=42,
-    label_dir="yolo_labels/", image_dir="images/",
+    output_dir="output/",
+    ratio=0.8,
+    seed=42,
+    label_dir="yolo_labels/",
+    image_dir="images/",
     class_file="classes.txt",
 )
 
 # Category filter (keep / remap categories per new classes.txt)
 filterer = FilterAnalyser(log_config=log_cfg)
 result = filterer.analyse(
-    "yolo_labels/", original_class_file="classes.txt",
-    new_class_file="classes_new.txt", output_dir="filtered/",
+    "yolo_labels/",
+    original_class_file="classes.txt",
+    new_class_file="classes_new.txt",
+    output_dir="filtered/",
 )
 
 # N-way partition (YOLO / LabelMe labels; images follow by stem)
 partitioner = PartitionAnalyser(log_config=log_cfg)
 result = partitioner.analyse(
-    output_dir="parts/", num=4,
-    label_dir="yolo_labels/", image_dir="images/",
+    output_dir="parts/",
+    num=4,
+    label_dir="yolo_labels/",
+    image_dir="images/",
 )
 
 # File sampling (labels, images, or both — random or sequential)
 sampler = SampleAnalyser(log_config=log_cfg)
 result = sampler.analyse(
-    output_dir="sampled/", count=10,
-    label_dir="yolo_labels/", shuffle=True, seed=42,
+    output_dir="sampled/",
+    count=10,
+    label_dir="yolo_labels/",
+    shuffle=True,
+    seed=42,
 )
 
 # ── Convert ──────────────────────────────────────────
@@ -285,22 +310,30 @@ result = sampler.analyse(
 log_cfg = LogConfig(name="convert", verbose=True)
 converter = YoloAndCocoConverter(source_to_target=True, log_config=log_cfg, strict_mode=True)
 result = converter.convert(
-    source_path="yolo_labels/", target_path="anno.json",
-    class_file="classes.txt", image_dir="images/",
+    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/",
+    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/", log_config=log_cfg,
+    label_dir="yolo_labels/",
+    image_dir="images/",
+    class_file="classes.txt",
+    is_show=True,
+    is_save=True,
+    output_dir="visualized/",
+    log_config=log_cfg,
 )
 result = visualizer.visualize()
 
@@ -370,7 +403,7 @@ For detailed developer guidance including advanced test commands, debugging, and
 
 ### 🧪 Testing
 
-**561 tests, 80% code coverage (5462 statements).**
+**606 tests, 80% code coverage (5532 statements).**
 
 ```bash
 pytest                                    # All tests
@@ -384,11 +417,11 @@ pytest tests/evaluate/test_evaluator.py     # Single module
 
 | Module | Coverage | Highlights |
 |--------|:--------:|------------|
-| `dataflow/label/` | 71% | models (84%), base (82%), utils (78%), coco_handler (74%), labelme_handler (71%), yolo_handler (61%) |
-| `dataflow/analyse/` | 84% | base (99%), log_templates (92%), sample (87%), utils (85%), split (85%), stats (83%), filter (76%), partition (74%) |
-| `dataflow/convert/` | 85% | labelme_and_yolo (93%), yolo_and_coco (89%), utils (87%), coco_and_labelme (86%), base (80%), rle (80%) |
+| `dataflow/label/` | 71% | models (84%), base (82%), utils (83%), coco_handler (74%), labelme_handler (71%), yolo_handler (61%) |
+| `dataflow/analyse/` | 84% | base (99%), log_templates (92%), sample (87%), utils (84%), split (85%), stats (82%), filter (76%), partition (74%) |
+| `dataflow/convert/` | 86% | labelme_and_yolo (93%), yolo_and_coco (89%), utils (88%), coco_and_labelme (88%), log_templates (85%), base (81%), rle (80%) |
 | `dataflow/visualize/` | 81% | yolo_vis (100%), labelme_vis (100%), coco_vis (93%), base (76%) |
-| `dataflow/evaluate/` | 87% | evaluator (100%), result (99%), metrics (93%), base (90%), utils (67%) |
+| `dataflow/evaluate/` | 90% | evaluator (100%), result (99%), metrics (93%), base (90%), utils (87%) |
 | `dataflow/cli/` | 74% | main (96%), visualize cmd (87%), utils (87%), evaluate cmd (83%), analyse cmd (65%), convert cmd (52%) |
 | `dataflow/util/` | 100% | logging (100%) |
 
@@ -397,11 +430,10 @@ pytest tests/evaluate/test_evaluator.py     # Single module
 ### 🎨 Code Quality
 
 ```bash
-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
+pip install -e .[dev]                       # Install dev dependencies
+ruff check dataflow tests samples           # Lint
+ruff format --check dataflow tests samples  # Format check
+mypy dataflow                               # Type check
 ```
 
 ### 🔗 Pre-commit Hooks (Optional)
@@ -411,7 +443,7 @@ pip install pre-commit
 pre-commit install            # Install git hooks (run once)
 
 # After this, every `git commit` auto-runs:
-#   black → isort → flake8 → whitespace checks
+#   ruff (lint, auto-fix) → ruff format → whitespace checks
 
 pre-commit run --all-files    # Manual run against all files
 ```
@@ -427,7 +459,7 @@ dataflow/
 ├── evaluate/        # pycocotools-based metrics, log templates
 ├── util/            # Unified logging (LogManager + format helpers)
 └── cli/             # CLI entry point, commands, validation
-tests/               # Unit & integration tests (561 tests, conftest fixtures)
+tests/               # Unit & integration tests (606 tests, conftest fixtures)
 samples/             # Python API usage examples (analyse, convert, visualize, evaluate, cli)
 assets/              # Test data (det/seg by format)
 specs/               # Canonical specifications (evaluate/ + formats/ + modules/)

{dataflow_cv-2.0.1 → dataflow_cv-3.0.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "dataflow-cv"
-version = "2.0.1"
+version = "3.0.0"
 description = "A computer vision dataset processing library — analyse, convert, visualize, and evaluate annotations across YOLO, LabelMe, and COCO formats"
 readme = "README.md"
 requires-python = ">=3.8"