lazylabel-gui 1.3.6__tar.gz → 1.3.7__tar.gz

lazylabel_gui-1.3.7/PKG-INFO

@@ -0,0 +1,227 @@
+Metadata-Version: 2.4
+Name: lazylabel-gui
+Version: 1.3.7
+Summary: An image segmentation GUI for generating ML ready mask tensors and annotations.
+Author-email: "Deniz N. Cakan" <deniz.n.cakan@gmail.com>
+License: MIT License
+        
+        Copyright (c) 2025 Deniz N. Cakan
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to whom the Software is
+        furnished to do so, subject to the following conditions:
+        
+        The above copyright notice and this permission notice shall be included in all
+        copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE.
+        
+Project-URL: Homepage, https://github.com/dnzckn/lazylabel
+Project-URL: Bug Tracker, https://github.com/dnzckn/lazylabel/issues
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Topic :: Scientific/Engineering :: Image Processing
+Classifier: Environment :: X11 Applications :: Qt
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: PyQt6>=6.9.0
+Requires-Dist: pyqtdarktheme==2.1.0
+Requires-Dist: torch>=2.7.1
+Requires-Dist: torchvision>=0.22.1
+Requires-Dist: segment-anything==1.0
+Requires-Dist: numpy>=2.1.2
+Requires-Dist: opencv-python>=4.11.0.86
+Requires-Dist: scipy>=1.15.3
+Requires-Dist: requests>=2.32.4
+Requires-Dist: tqdm>=4.67.1
+Provides-Extra: dev
+Requires-Dist: pytest>=7.0.0; extra == "dev"
+Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
+Requires-Dist: pytest-mock>=3.10.0; extra == "dev"
+Requires-Dist: pytest-qt>=4.2.0; extra == "dev"
+Requires-Dist: ruff>=0.8.0; extra == "dev"
+Dynamic: license-file
+
+# LazyLabel
+
+<div align="center">
+  <img src="https://raw.githubusercontent.com/dnzckn/LazyLabel/main/src/lazylabel/demo_pictures/logo2.png" alt="LazyLabel Logo" style="height:60px; vertical-align:middle;" />
+  <img src="https://raw.githubusercontent.com/dnzckn/LazyLabel/main/src/lazylabel/demo_pictures/logo_black.png" alt="LazyLabel Cursive" style="height:60px; vertical-align:middle;" />
+</div>
+
+**AI-Assisted Image Segmentation for Machine Learning Dataset Preparation**
+
+LazyLabel combines Meta's Segment Anything Model (SAM) with comprehensive manual annotation tools to accelerate the creation of pixel-perfect segmentation masks for computer vision applications.
+
+<div align="center">
+  <img src="https://raw.githubusercontent.com/dnzckn/LazyLabel/main/src/lazylabel/demo_pictures/gui.PNG" alt="LazyLabel Screenshot" width="800"/>
+</div>
+
+---
+
+## Quick Start
+
+```bash
+pip install lazylabel-gui
+lazylabel-gui
+```
+
+**From source:**
+```bash
+git clone https://github.com/dnzckn/LazyLabel.git
+cd LazyLabel
+pip install -e .
+lazylabel-gui
+```
+
+**Requirements:** Python 3.10+, 8GB RAM, ~2.5GB disk space (for model weights)
+
+---
+
+## Core Features
+
+### AI-Powered Segmentation
+LazyLabel leverages Meta's SAM for intelligent object detection:
+- Single-click object segmentation
+- Interactive refinement with positive/negative points  
+- Support for both SAM 1.0 and SAM 2.1 models
+- GPU acceleration with automatic CPU fallback
+
+### Manual Annotation Tools
+When precision matters:
+- Polygon drawing with vertex-level editing
+- Bounding box annotations for object detection
+- Edit mode for adjusting existing segments
+- Merge tool for combining related segments
+
+### Image Processing & Filtering
+Advanced preprocessing capabilities:
+- **FFT filtering**: Remove noise and enhance edges
+- **Channel thresholding**: Isolate objects by color
+- **Border cropping**: Define crop regions that set pixels outside the area to zero in saved outputs
+- **View adjustments**: Brightness, contrast, gamma correction
+
+### Multi-View Mode
+Process multiple images efficiently:
+- Annotate up to 4 images simultaneously
+- Synchronized zoom and pan across views
+- Mirror annotations to all linked images
+
+---
+
+## Export Formats
+
+### NPZ Format (Semantic Segmentation)
+One-hot encoded masks optimized for deep learning:
+
+```python
+import numpy as np
+
+data = np.load('image.npz')
+mask = data['mask']  # Shape: (height, width, num_classes)
+
+# Each channel represents one class
+sky = mask[:, :, 0]
+boats = mask[:, :, 1]
+cats = mask[:, :, 2]
+dogs = mask[:, :, 3]
+```
+
+### YOLO Format (Object Detection)
+Normalized polygon coordinates for YOLO training:
+```
+0 0.234 0.456 0.289 0.478 0.301 0.523 ...
+1 0.567 0.123 0.598 0.145 0.612 0.189 ...
+```
+
+### Class Aliases (JSON)
+Maintains consistent class naming across datasets:
+```json
+{
+  "0": "background",
+  "1": "person",
+  "2": "vehicle"
+}
+```
+
+---
+
+## Typical Workflow
+
+1. **Open folder** containing your images
+2. **Click objects** to generate AI masks (mode 1)
+3. **Refine** with additional points or manual tools
+4. **Assign classes** and organize in the class table
+5. **Export** as NPZ or YOLO format
+
+### Advanced Preprocessing Workflow
+
+For challenging images:
+1. Apply **FFT filtering** to reduce noise
+2. Use **channel thresholding** to isolate color ranges
+3. Enable **"Operate on View"** to pass filtered images to SAM
+4. Fine-tune with manual tools
+
+---
+
+## Advanced Features
+
+### Multi-View Mode
+Access via the "Multi" tab to process multiple images:
+- 2-view (side-by-side) or 4-view (grid) layouts
+- Annotations mirror across linked views automatically
+- Synchronized zoom maintains alignment
+
+### SAM 2.1 Support
+LazyLabel supports both SAM 1.0 (default) and SAM 2.1 models. SAM 2.1 offers improved segmentation accuracy and better handling of complex boundaries.
+
+To use SAM 2.1 models:
+1. Install the SAM 2 package:
+   ```bash
+   pip install git+https://github.com/facebookresearch/sam2.git
+   ```
+2. Download a SAM 2.1 model (e.g., `sam2.1_hiera_large.pt`) from the [SAM 2 repository](https://github.com/facebookresearch/sam2)
+3. Place the model file in LazyLabel's models folder:
+   - If installed via pip: `~/.local/share/lazylabel/models/` (or equivalent on your system)
+   - If running from source: `src/lazylabel/models/`
+4. Select the SAM 2.1 model from the dropdown in LazyLabel's settings
+
+Note: SAM 1.0 models are automatically downloaded on first use.
+
+---
+
+## Key Shortcuts
+
+| Action | Key | Description |
+|--------|-----|-------------|
+| AI Mode | `1` | SAM point-click segmentation |
+| Draw Mode | `2` | Manual polygon creation |
+| Edit Mode | `E` | Modify existing segments |
+| Accept AI Segment | `Space` | Confirm AI segment suggestion |
+| Save | `Enter` | Save annotations |
+| Merge | `M` | Combine selected segments |
+| Pan Mode | `Q` | Enter pan mode |
+| Pan | `WASD` | Navigate image |
+| Delete | `V`/`Delete` | Remove segments |
+| Undo/Redo | `Ctrl+Z/Y` | Action history |
+
+---
+
+## Documentation
+
+- [Usage Manual](src/lazylabel/USAGE_MANUAL.md) - Comprehensive feature guide
+- [Architecture Guide](src/lazylabel/ARCHITECTURE.md) - Technical implementation details
+- [GitHub Issues](https://github.com/dnzckn/LazyLabel/issues) - Report bugs or request features
+
+---

lazylabel_gui-1.3.7/README.md

@@ -0,0 +1,172 @@
+# LazyLabel
+
+<div align="center">
+  <img src="https://raw.githubusercontent.com/dnzckn/LazyLabel/main/src/lazylabel/demo_pictures/logo2.png" alt="LazyLabel Logo" style="height:60px; vertical-align:middle;" />
+  <img src="https://raw.githubusercontent.com/dnzckn/LazyLabel/main/src/lazylabel/demo_pictures/logo_black.png" alt="LazyLabel Cursive" style="height:60px; vertical-align:middle;" />
+</div>
+
+**AI-Assisted Image Segmentation for Machine Learning Dataset Preparation**
+
+LazyLabel combines Meta's Segment Anything Model (SAM) with comprehensive manual annotation tools to accelerate the creation of pixel-perfect segmentation masks for computer vision applications.
+
+<div align="center">
+  <img src="https://raw.githubusercontent.com/dnzckn/LazyLabel/main/src/lazylabel/demo_pictures/gui.PNG" alt="LazyLabel Screenshot" width="800"/>
+</div>
+
+---
+
+## Quick Start
+
+```bash
+pip install lazylabel-gui
+lazylabel-gui
+```
+
+**From source:**
+```bash
+git clone https://github.com/dnzckn/LazyLabel.git
+cd LazyLabel
+pip install -e .
+lazylabel-gui
+```
+
+**Requirements:** Python 3.10+, 8GB RAM, ~2.5GB disk space (for model weights)
+
+---
+
+## Core Features
+
+### AI-Powered Segmentation
+LazyLabel leverages Meta's SAM for intelligent object detection:
+- Single-click object segmentation
+- Interactive refinement with positive/negative points  
+- Support for both SAM 1.0 and SAM 2.1 models
+- GPU acceleration with automatic CPU fallback
+
+### Manual Annotation Tools
+When precision matters:
+- Polygon drawing with vertex-level editing
+- Bounding box annotations for object detection
+- Edit mode for adjusting existing segments
+- Merge tool for combining related segments
+
+### Image Processing & Filtering
+Advanced preprocessing capabilities:
+- **FFT filtering**: Remove noise and enhance edges
+- **Channel thresholding**: Isolate objects by color
+- **Border cropping**: Define crop regions that set pixels outside the area to zero in saved outputs
+- **View adjustments**: Brightness, contrast, gamma correction
+
+### Multi-View Mode
+Process multiple images efficiently:
+- Annotate up to 4 images simultaneously
+- Synchronized zoom and pan across views
+- Mirror annotations to all linked images
+
+---
+
+## Export Formats
+
+### NPZ Format (Semantic Segmentation)
+One-hot encoded masks optimized for deep learning:
+
+```python
+import numpy as np
+
+data = np.load('image.npz')
+mask = data['mask']  # Shape: (height, width, num_classes)
+
+# Each channel represents one class
+sky = mask[:, :, 0]
+boats = mask[:, :, 1]
+cats = mask[:, :, 2]
+dogs = mask[:, :, 3]
+```
+
+### YOLO Format (Object Detection)
+Normalized polygon coordinates for YOLO training:
+```
+0 0.234 0.456 0.289 0.478 0.301 0.523 ...
+1 0.567 0.123 0.598 0.145 0.612 0.189 ...
+```
+
+### Class Aliases (JSON)
+Maintains consistent class naming across datasets:
+```json
+{
+  "0": "background",
+  "1": "person",
+  "2": "vehicle"
+}
+```
+
+---
+
+## Typical Workflow
+
+1. **Open folder** containing your images
+2. **Click objects** to generate AI masks (mode 1)
+3. **Refine** with additional points or manual tools
+4. **Assign classes** and organize in the class table
+5. **Export** as NPZ or YOLO format
+
+### Advanced Preprocessing Workflow
+
+For challenging images:
+1. Apply **FFT filtering** to reduce noise
+2. Use **channel thresholding** to isolate color ranges
+3. Enable **"Operate on View"** to pass filtered images to SAM
+4. Fine-tune with manual tools
+
+---
+
+## Advanced Features
+
+### Multi-View Mode
+Access via the "Multi" tab to process multiple images:
+- 2-view (side-by-side) or 4-view (grid) layouts
+- Annotations mirror across linked views automatically
+- Synchronized zoom maintains alignment
+
+### SAM 2.1 Support
+LazyLabel supports both SAM 1.0 (default) and SAM 2.1 models. SAM 2.1 offers improved segmentation accuracy and better handling of complex boundaries.
+
+To use SAM 2.1 models:
+1. Install the SAM 2 package:
+   ```bash
+   pip install git+https://github.com/facebookresearch/sam2.git
+   ```
+2. Download a SAM 2.1 model (e.g., `sam2.1_hiera_large.pt`) from the [SAM 2 repository](https://github.com/facebookresearch/sam2)
+3. Place the model file in LazyLabel's models folder:
+   - If installed via pip: `~/.local/share/lazylabel/models/` (or equivalent on your system)
+   - If running from source: `src/lazylabel/models/`
+4. Select the SAM 2.1 model from the dropdown in LazyLabel's settings
+
+Note: SAM 1.0 models are automatically downloaded on first use.
+
+---
+
+## Key Shortcuts
+
+| Action | Key | Description |
+|--------|-----|-------------|
+| AI Mode | `1` | SAM point-click segmentation |
+| Draw Mode | `2` | Manual polygon creation |
+| Edit Mode | `E` | Modify existing segments |
+| Accept AI Segment | `Space` | Confirm AI segment suggestion |
+| Save | `Enter` | Save annotations |
+| Merge | `M` | Combine selected segments |
+| Pan Mode | `Q` | Enter pan mode |
+| Pan | `WASD` | Navigate image |
+| Delete | `V`/`Delete` | Remove segments |
+| Undo/Redo | `Ctrl+Z/Y` | Action history |
+
+---
+
+## Documentation
+
+- [Usage Manual](src/lazylabel/USAGE_MANUAL.md) - Comprehensive feature guide
+- [Architecture Guide](src/lazylabel/ARCHITECTURE.md) - Technical implementation details
+- [GitHub Issues](https://github.com/dnzckn/LazyLabel/issues) - Report bugs or request features
+
+---

{lazylabel_gui-1.3.6 → lazylabel_gui-1.3.7}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "lazylabel-gui"
-version = "1.3.6"
+version = "1.3.7"
 authors = [
   { name="Deniz N. Cakan", email="deniz.n.cakan@gmail.com" },
 ]

{lazylabel_gui-1.3.6 → lazylabel_gui-1.3.7}/src/lazylabel/ui/main_window.py

@@ -1582,6 +1582,19 @@ class MainWindow(QMainWindow):
         # Reset AI mode state
         self.ai_click_start_pos = None
 
+        # Reset all link states to linked when navigating to new images
+        config = self._get_multi_view_config()
+        num_viewers = config["num_viewers"]
+        self.multi_view_linked = [True] * num_viewers
+
+        # Reset unlink button appearances to default linked state
+        if hasattr(self, "multi_view_unlink_buttons"):
+            for i, button in enumerate(self.multi_view_unlink_buttons):
+                if i < num_viewers:
+                    button.setText("X")
+                    button.setToolTip("Unlink this image from mirroring")
+                    button.setStyleSheet("")
+
         # Update UI lists to reflect cleared state
         self._update_all_lists()
 
@@ -2829,7 +2842,8 @@ class MainWindow(QMainWindow):
         for i in range(num_viewers):
             image_path = self.multi_view_images[i]
 
-            if not image_path:
+            # Skip if no image or if this viewer is unlinked
+            if not image_path or not self.multi_view_linked[i]:
                 continue
 
             # Filter segments for this viewer
@@ -6802,21 +6816,20 @@ class MainWindow(QMainWindow):
 
     def _toggle_multi_view_link(self, image_index):
         """Toggle the link status for a specific image in multi-view."""
-        if 0 <= image_index < 2:
-            self.multi_view_linked[image_index] = not self.multi_view_linked[
-                image_index
-            ]
+        # Check bounds and link status - only allow unlinking when currently linked
+        if (
+            0 <= image_index < len(self.multi_view_linked)
+            and self.multi_view_linked[image_index]
+        ):
+            # Currently linked - allow unlinking
+            self.multi_view_linked[image_index] = False
 
-            # Update button appearance
+            # Update button appearance to show unlinked state
             button = self.multi_view_unlink_buttons[image_index]
-            if self.multi_view_linked[image_index]:
-                button.setText("X")
-                button.setToolTip("Unlink this image from mirroring")
-                button.setStyleSheet("")
-            else:
-                button.setText("↪")
-                button.setToolTip("Link this image to mirroring")
-                button.setStyleSheet("background-color: #ffcccc;")
+            button.setText("↪")
+            button.setToolTip("This image is unlinked from mirroring")
+            button.setStyleSheet("background-color: #ff4444; color: white;")
+        # If already unlinked or invalid index, do nothing (prevent re-linking)
 
     def _start_background_image_discovery(self):
         """Start background discovery of all image files."""

{lazylabel_gui-1.3.6 → lazylabel_gui-1.3.7}/src/lazylabel/ui/modes/multi_view_mode.py

@@ -369,12 +369,20 @@ class MultiViewModeHandler(BaseModeHandler):
 
         paired_segment = {"type": "Polygon", "views": {}}
 
-        # Add view data for all viewers with same coordinates
+        # Add view data for current viewer and mirror to linked viewers only
+        paired_segment["views"][viewer_index] = view_data
+
+        # Mirror to other viewers only if they are linked
         for viewer_idx in range(num_viewers):
-            paired_segment["views"][viewer_idx] = {
-                "vertices": view_data["vertices"].copy(),
-                "mask": None,
-            }
+            if (
+                viewer_idx != viewer_index
+                and self.main_window.multi_view_linked[viewer_idx]
+                and self.main_window.multi_view_images[viewer_idx] is not None
+            ):
+                paired_segment["views"][viewer_idx] = {
+                    "vertices": view_data["vertices"].copy(),
+                    "mask": None,
+                }
 
         # Add to segment manager
         self.main_window.segment_manager.add_segment(paired_segment)
@@ -783,9 +791,13 @@ class MultiViewModeHandler(BaseModeHandler):
         # Add the current viewer's data
         paired_segment["views"][viewer_index] = view_data
 
-        # Mirror to all other viewers with same coordinates (they should align between linked images)
+        # Mirror to all other viewers with same coordinates (only if they are linked)
         for other_viewer_index in range(num_viewers):
-            if other_viewer_index != viewer_index:
+            if (
+                other_viewer_index != viewer_index
+                and self.main_window.multi_view_linked[other_viewer_index]
+                and self.main_window.multi_view_images[other_viewer_index] is not None
+            ):
                 mirrored_view_data = {
                     "vertices": view_data[
                         "vertices"
@@ -807,11 +819,27 @@ class MultiViewModeHandler(BaseModeHandler):
 
         # Update UI
         self.main_window._update_all_lists()
-        viewer_count_text = "all viewers" if num_viewers > 2 else "both viewers"
-        self.main_window._show_notification(
-            f"Polygon created and mirrored to {viewer_count_text}."
+
+        # Count linked viewers (excluding the source viewer)
+        linked_viewers_count = sum(
+            1
+            for i in range(num_viewers)
+            if i != viewer_index
+            and self.main_window.multi_view_linked[i]
+            and self.main_window.multi_view_images[i] is not None
         )
 
+        if linked_viewers_count == 0:
+            viewer_count_text = "created (no linked viewers to mirror to)"
+        elif linked_viewers_count == 1:
+            viewer_count_text = "created and mirrored to 1 linked viewer"
+        else:
+            viewer_count_text = (
+                f"created and mirrored to {linked_viewers_count} linked viewers"
+            )
+
+        self.main_window._show_notification(f"Polygon {viewer_count_text}.")
+
         # Clear polygon state for this viewer
         self._clear_multi_view_polygon(viewer_index)