Fix sam2

.github/workflows/tests.yml

@@ -91,7 +91,7 @@ jobs:
         run: docker compose -f label_studio_ml/examples/${{ matrix.backend_dir_name }}/docker-compose.yml up -d --build
 
       - name: Wait for stack
-        timeout-minutes: 10
+        timeout-minutes: 20
         run: |
           while [ "$(curl -s -o /dev/null -L -w ''%{http_code}'' "http://localhost:9090/health")" != "200" ]; do
             echo "=> Waiting for service to become available" && sleep 2s

label_studio_ml/api.py

@@ -56,7 +56,7 @@ def _predict():
     data = request.json
     tasks = data.get('tasks')
     label_config = data.get('label_config')
-    project = data.get('project')
+    project = str(data.get('project'))
     project_id = project.split('.', 1)[0] if project else None
     params = data.get('params', {})
     context = params.pop('context', {})
@@ -123,8 +123,14 @@ def webhook():
     project_id = str(data['project']['id'])
     label_config = data['project']['label_config']
     model = MODEL_CLASS(project_id, label_config=label_config)
-    model.fit(event, data)
-    return jsonify({}), 201
+    result = model.fit(event, data)
+
+    try:
+        response = jsonify({'result': result, 'status': 'ok'})
+    except Exception as e:
+        response = jsonify({'error': str(e), 'status': 'error'})
+
+    return response, 201
 
 
 @_server.route('/health', methods=['GET'])

label_studio_ml/default_configs/_wsgi.py.tmpl

@@ -4,29 +4,33 @@ import json
 import logging
 import logging.config
 
-logging.config.dictConfig({{
-  "version": 1,
-  "formatters": {{
-    "standard": {{
-      "format": "[%(asctime)s] [%(levelname)s] [%(name)s::%(funcName)s::%(lineno)d] %(message)s"
-    }}
-  }},
-  "handlers": {{
-    "console": {{
-      "class": "logging.StreamHandler",
-      "level": os.getenv('LOG_LEVEL'),
-      "stream": "ext://sys.stdout",
-      "formatter": "standard"
-    }}
-  }},
-  "root": {{
-    "level": os.getenv('LOG_LEVEL'),
-    "handlers": [
-      "console"
-    ],
-    "propagate": True
-  }}
-}})
+# Set a default log level if LOG_LEVEL is not defined
+log_level = os.getenv("LOG_LEVEL", "INFO")
+
+logging.config.dictConfig(
+    {
+        "version": 1,
+        "disable_existing_loggers": False,  # Prevent overriding existing loggers
+        "formatters": {
+            "standard": {
+                "format": "[%(asctime)s] [%(levelname)s] [%(name)s::%(funcName)s::%(lineno)d] %(message)s"
+            }
+        },
+        "handlers": {
+            "console": {
+                "class": "logging.StreamHandler",
+                "level": log_level,
+                "stream": "ext://sys.stdout",
+                "formatter": "standard",
+            }
+        },
+        "root": {
+            "level": log_level,
+            "handlers": ["console"],
+            "propagate": True,
+        },
+    }
+)
 
 from label_studio_ml.api import init_app
 from {script} import {model_class}
@@ -59,7 +63,7 @@ if __name__ == "__main__":
         '-d', '--debug', dest='debug', action='store_true',
         help='Switch debug mode')
     parser.add_argument(
-        '--log-level', dest='log_level', choices=['DEBUG', 'INFO', 'WARNING', 'ERROR'], default=None,
+        '--log-level', dest='log_level', choices=['DEBUG', 'INFO', 'WARNING', 'ERROR'], default=log_level,
         help='Logging level')
     parser.add_argument(
         '--model-dir', dest='model_dir', default=os.path.dirname(__file__),

label_studio_ml/examples/segment_anything_2_image/README.md

@@ -133,15 +133,29 @@ cd label_studio_ml/examples/segment_anything_2_image
 pip install -r requirements.txt
 ```
 
-2. Download [`segment-anything-2` repo](https://github.com/facebookresearch/segment-anything-2) into the root directory. Install SegmentAnything model and download checkpoints using [the official Meta documentation](https://github.com/facebookresearch/segment-anything-2?tab=readme-ov-file#installation)
+2. Download [`segment-anything-2` repo](https://github.com/facebookresearch/segment-anything-2) into the ROOT directory. Install SegmentAnything model and download checkpoints using [the official Meta documentation](https://github.com/facebookresearch/segment-anything-2?tab=readme-ov-file#installation)
 
+You should now have the following folder structure: 
+
+ROOT_DIR (i.e. ~)
+
+    | label-studio-ml-backend 
+        | label-studio-ml 
+            | examples 
+                | segment_anything_2_image
+    | sam2
+        | sam2 
+        | checkpoints 
 
 3. Then you can start the ML backend on the default port `9090`:
 
 ```bash
-cd ../
-label-studio-ml start ./segment_anything_2_image
+cd ~/sam2
+label-studio-ml start ../label-studio-ml-backend/label-studio-ml/examples/segement_anything_2_image
 ```
+Make sure that you run this code from the sam2 directory in your root directory, as Meta has added a check that makes other folder structures break. Additionally, we rely on a `os.getcwd()` command to get the rest of the code working properly. 
+
+To change the port that you're running on, simply add  `--p <port number>` to the end of the above command.
 
 4. Connect running ML backend server to Label Studio: go to your project `Settings -> Machine Learning -> Add Model` and specify `http://localhost:9090` as a URL. Read more in the official [Label Studio documentation](https://labelstud.io/guide/ml#Connect-the-model-to-Label-Studio).
 

label_studio_ml/examples/segment_anything_2_image/model.py

@@ -1,6 +1,7 @@
 import torch
 import numpy as np
 import os
+import sys
 import pathlib
 from typing import List, Dict, Optional
 from uuid import uuid4
@@ -9,6 +10,10 @@
 from label_studio_sdk.converter import brush
 from label_studio_sdk._extensions.label_studio_tools.core.utils.io import get_local_path
 from PIL import Image
+
+ROOT_DIR = os.getcwd()
+sys.path.insert(0, ROOT_DIR)
+
 from sam2.build_sam import build_sam2
 from sam2.sam2_image_predictor import SAM2ImagePredictor
 
@@ -29,7 +34,7 @@
 
 
 # build path to the model checkpoint
-sam2_checkpoint = str(pathlib.Path(__file__).parent / SEGMENT_ANYTHING_2_REPO_PATH / "checkpoints" / MODEL_CHECKPOINT)
+sam2_checkpoint = str(os.path.join(ROOT_DIR, "checkpoints", MODEL_CHECKPOINT))
 
 sam2_model = build_sam2(MODEL_CONFIG, sam2_checkpoint, device=DEVICE)
 
@@ -152,5 +157,5 @@ def predict(self, tasks: List[Dict], context: Optional[Dict] = None, **kwargs) -
             from_name=from_name,
             to_name=to_name,
             label=selected_label)
-        
+
         return ModelResponse(predictions=predictions)

label_studio_ml/examples/yolo/.dockerignore

@@ -11,6 +11,7 @@
 !tests/*
 !control_models/*
 !models/*
+!utils/*
 
 # Include requirements files
 !requirements*.txt

label_studio_ml/examples/yolo/Dockerfile

@@ -49,14 +49,14 @@ WORKDIR /app
 
 COPY . ./
 
-WORKDIR /app/models
-
 # Download the YOLO models
-RUN yolo predict model=yolov8m.pt source=/app/tests/car.jpg \
-    && yolo predict model=yolov8n.pt source=/app/tests/car.jpg \
-    && yolo predict model=yolov8n-cls.pt source=/app/tests/car.jpg \
-    && yolo predict model=yolov8n-seg.pt source=/app/tests/car.jpg
-
-WORKDIR /app
+RUN /bin/sh -c 'if [ ! -f /app/models/yolov8m.pt ]; then \
+    yolo predict model=/app/models/yolov8m.pt source=/app/tests/car.jpg \
+    && yolo predict model=/app/models/yolov8n.pt source=/app/tests/car.jpg \
+    && yolo predict model=/app/models/yolov8n-cls.pt source=/app/tests/car.jpg \
+    && yolo predict model=/app/models/yolov8n-seg.pt source=/app/tests/car.jpg; \
+    fi'
+
+ENV PYTHONPATH=/app
 
 CMD ["/app/start.sh"]

label_studio_ml/examples/yolo/README.md

@@ -36,16 +36,17 @@ making it easier to annotate large datasets and ensure high-quality predictions.
 
 **Supported Features**
 
-| YOLO Task Name                        | LS Control Tag                       | Prediction Supported | LS Import Supported | LS Export Supported |
-|---------------------------------------|--------------------------------------|----------------------|---------------------|---------------------|
-| Object Detection                      | `<RectangleLabels>`                  | ✅                    | YOLO, COCO          | YOLO, COCO          |
-| Oriented Bounding Boxes (OBB)         | `<RectangleLabels model_obb="true">` | ✅                    | YOLO                | YOLO                |
-| Image Instance Segmentation: Polygons | `<PolygonLabels>`                    | ✅                    | COCO                | YOLO, COCO          |
-| Image Semantic Segmentation: Masks    | `<BrushLabels>`                      | ❌                    | Native              | Native              |
-| Image Classification                  | `<Choices>`                          | ✅                    | Native              | Native              |
-| Pose Detection                        | `<KeyPoints>`                        | ✅                    | Native              | Native              |
-| Video Object Tracking                 | `<VideoRectangle>`                   | ✅                    | Native              | Native              |
-| Video Temporal Classification         | `<TimelineLabels>`                   | Coming soon          | Native              | Native              |
+| YOLO Task Name                                               | LS Control Tag                       | Prediction Supported | LS Import Supported | LS Export Supported |
+|--------------------------------------------------------------|--------------------------------------|----------------------|---------------------|---------------------|
+| Object Detection                                             | `<RectangleLabels>`                  | ✅                    | YOLO, COCO          | YOLO, COCO          |
+| Oriented Bounding Boxes (OBB)                                | `<RectangleLabels model_obb="true">` | ✅                    | YOLO                | YOLO                |
+| Image Instance Segmentation: Polygons                        | `<PolygonLabels>`                    | ✅                    | COCO                | YOLO, COCO          |
+| Image Semantic Segmentation: Masks                           | `<BrushLabels>`                      | ❌                    | Native              | Native              |
+| Image Classification                                         | `<Choices>`                          | ✅                    | Native              | Native              |
+| Pose Detection                                               | `<KeyPoints>`                        | ✅                    | Native              | Native              |
+| Video Object Tracking                                        | `<VideoRectangle>`                   | ✅                    | Native              | Native              |
+| [Video Temporal Classification](./README_TIMELINE_LABELS.md) | `<TimelineLabels>`                   | ✅                    | Native              | Native              |
+
 
 * **LS Control Tag**: Label Studio [control tag](https://labelstud.io/tags/) from the labeling configuration. 
 * **LS Import Supported**: Indicates whether Label Studio supports Import from YOLO format to Label Studio (using the LS converter).
@@ -82,7 +83,7 @@ This tutorial uses the [YOLO example](https://github.com/HumanSignal/label-studi
 
 4. Then from the **Model** page in the project settings, [connect the model](https://labelstud.io/guide/ml#Connect-the-model-to-Label-Studio). The default URL is `http://localhost:9090`. 
 
-5. Add images to Label Studio.
+5. Add images or video (depending on tasks you are going to solve) to Label Studio.
 
 6. Open any task in the Data Manager and see the predictions from the YOLO model.
 
@@ -97,11 +98,13 @@ This tutorial uses the [YOLO example](https://github.com/HumanSignal/label-studi
 
 **Control tags**
 
+- `<Choices>` - [Classification](https://labelstud.io/tags/choices); image classification task
 - `<RectangleLabels>` - [Bounding boxes](https://labelstud.io/tags/rectanglelabels); object detection task
 - `<PolygonLabels>` - [Polygons](https://labelstud.io/tags/polygonlables); segmentation task
-- `<VideoRectangle>` - [Video bounding boxes](https://labelstud.io/tags/videorectangle); video object tracking task
+- `<VideoRectangle>` - [Video bounding boxes](https://labelstud.io/tags/videorectangle); object tracking task for videos
 - `<KeyPointLabels>` - [Key points](https://labelstud.io/tags/keypointlabels); pose detection task
-- `<Choices>` - [Classification](https://labelstud.io/tags/choices)
+- `<TimelineLabels>` - [Temporal labels for videos](https://labelstud.io/tags/timelinelabels); multi-label temporal classification task for videos
+
 
 **How to skip the control tag?**
 
@@ -681,6 +684,47 @@ Small models like `yolov8n.pt` are recommended for real-time tracking, however,
 
 <br>
 
+
+## Video temporal classification using `TimelineLabels`
+
+This ML backend supports temporal multi-label video classification for the `TimelineLabels` control tag in Label Studio. 
+There are two modes available:
+- **Simple:** In the simple mode, the model uses pre-trained YOLO classes to generate predictions without additional training.  
+- **Trainable:** In the [trainable mode](README_TIMELINE_LABELS.md), the model can be trained on custom labels and annotations submitted in Label Studio like a few shot learning way
+when training is performed on a small number of annotations on the fly.  
+
+### Labeling config
+
+```xml
+<View>
+  <Video name="video" value="$video"/>
+  <TimelineLabels 
+          name="label" toName="video" 
+          model_trainable="false" model_score_threshold="0.25">
+    <Label value="Ball" predicted_values="soccer_ball" />
+    <Label value="tiger_shark" />
+  </TimelineLabels>
+</View>
+```
+
+### Model training
+
+For more details on using the `TimelineLabels` ML backend, including training the model 
+and adjusting neural network classifier parameters, please refer to 
+**[README_TIMELINE_LABELS.md](README_TIMELINE_LABELS.md)**.
+
+### Default model
+
+`yolov8n-cls.pt` is the default classification model for simple mode.
+
+
+<br>
+
+-------------------
+
+<br>
+
+
 ## Run the YOLO ML backend
 
 

label_studio_ml/examples/yolo/README_DEVELOP.md

@@ -47,13 +47,19 @@ classDiagram
         +create_video_rectangles(results, path) List[Dict]
         +update_tracker_params(yaml_path: str, prefix: str) str | None
     }
+
+    class TimelineLabelsModel {
+        +predict_regions(path: str) List[Dict]
+        +fit(event, data)
+    }
 
     ControlModel <|-- RectangleLabelsModel
     ControlModel <|-- RectangleLabelsObbModel
     ControlModel <|-- PolygonLabelsModel
     ControlModel <|-- ChoicesModel
     ControlModel <|-- KeyPointLabelsModel
     ControlModel <|-- VideoRectangleModel
+    ControlModel <|-- TimelineLabelsModel
 
 ```
 
@@ -122,6 +128,12 @@ The architecture of the project is modular and is primarily centered around inte
      - `create_video_rectangles()`: Processes the output of the tracking model to create a sequence of bounding boxes across video frames.
      - `update_tracker_params()`: Customizes the tracking parameters based on settings in Label Studio’s configuration.
 
+8. **`control_models/timelinelabels.py` (TimelineLabelsModel)**:
+    - **Purpose**: Supports the training of a YOLO model on video data incrementally by updating the model with new annotations as they are submitted.
+    - **Key Functions**:
+      - `predict_regions()`: Runs YOLO on video frames and returns the predictions.
+      - `fit()`: Placeholder method for updating the model with new annotations.
+
 ### **Module Interaction**
 
 - **Workflow**: The main workflow begins with `model.py`, which reads tasks and the Label Studio configuration to detect and instantiate the appropriate control models. These control models are responsible for making predictions using the YOLO model and converting the results into a format that Label Studio can use for annotations.