ovkit 0.2.0__tar.gz → 0.3.0__tar.gz

ovkit-0.2.0/README.md → ovkit-0.3.0/PKG-INFO

@@ -1,3 +1,59 @@
+Metadata-Version: 2.5
+Name: ovkit
+Version: 0.3.0
+Summary: A simple Python inference API for OpenVINO: one import, one Model class, clean Results — with AUTO/NPU, async, and INT8.
+Project-URL: Homepage, https://github.com/leeyunjai82/ovkit
+Project-URL: Documentation, https://leeyunjai82.github.io/ovkit/
+Project-URL: Repository, https://github.com/leeyunjai82/ovkit
+Project-URL: Issues, https://github.com/leeyunjai82/ovkit/issues
+Project-URL: Models, https://huggingface.co/leeyunjai/ovkit-models
+Author: ovkit contributors
+License: Apache-2.0
+License-File: LICENSE
+Keywords: computer-vision,detection,inference,npu,openvino,quantization
+Classifier: Development Status :: 3 - Alpha
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.10
+Requires-Dist: huggingface-hub>=0.20
+Requires-Dist: numpy>=1.21
+Requires-Dist: opencv-python-headless>=4.5
+Requires-Dist: openvino>=2024.0
+Requires-Dist: pillow>=9.0
+Requires-Dist: pyyaml>=6.0
+Provides-Extra: all
+Requires-Dist: anomalib; extra == 'all'
+Requires-Dist: nncf; extra == 'all'
+Requires-Dist: onnx; extra == 'all'
+Requires-Dist: openvino-genai; extra == 'all'
+Requires-Dist: optimum-intel; extra == 'all'
+Requires-Dist: scipy; extra == 'all'
+Requires-Dist: torch>=2.0; extra == 'all'
+Requires-Dist: torchvision>=0.15; extra == 'all'
+Provides-Extra: anomaly
+Requires-Dist: anomalib; extra == 'anomaly'
+Provides-Extra: dev
+Requires-Dist: black; extra == 'dev'
+Requires-Dist: pytest>=7; extra == 'dev'
+Requires-Dist: ruff; extra == 'dev'
+Provides-Extra: genai
+Requires-Dist: openvino-genai; extra == 'genai'
+Requires-Dist: optimum-intel; extra == 'genai'
+Provides-Extra: hand
+Requires-Dist: mediapipe; extra == 'hand'
+Provides-Extra: quant
+Requires-Dist: nncf; extra == 'quant'
+Provides-Extra: train
+Requires-Dist: onnx; extra == 'train'
+Requires-Dist: scipy; extra == 'train'
+Requires-Dist: torch>=2.0; extra == 'train'
+Requires-Dist: torchvision>=0.15; extra == 'train'
+Description-Content-Type: text/markdown
+
 <div align="center">
 
 <img src="https://raw.githubusercontent.com/leeyunjai82/ovkit/main/docs/_static/logo.svg" width="110" alt="ovkit logo"/>
@@ -93,6 +149,11 @@ camera index).
 | `attention` | `1 person looking at: laptop` | gaze + object detection (ray-cast into the boxes) |
 | `anonymize` | the picture with every face pixelated | face (and plate) detection + redaction |
 | `face_match` | `('yunjai', 0.81)` — who this is | embedding + cosine matching against your gallery |
+| `count` | `pencil 3 · cup 1` (optionally one kind only) | detection + per-kind tally |
+| `posture` | `neck 32° — sit up (6s)` | pose + neck angle **over time** |
+| `exercise` | `squat x 12 (down)` | pose + joint-angle hysteresis (squat, push-up) |
+| `attendance` | `present 24/26` + `roll.csv` | face detection + roster matching |
+| `teach` | your own classes, from example photos | embedding + cosine k-NN (5 modes) |
 
 ```python
 from ovkit import Model, list_pipelines
@@ -104,6 +165,45 @@ Model("face_analyze", attributes=("age_gender",))   # configure what runs
 ```
 
 Aliases: `ocr`, `anpr`, `blur`, `driver`, `describe`, `faces`, `people`, `vehicle`, `tracking`, `reid`.
+
+## Teach your own AI (no GPU)
+
+A Teachable Machine in five lines — an embedding model turns examples into
+vectors, new inputs match the nearest ones. Five modes: `photo` (default),
+`face` (expressions), `hand` (needs `ovkit[hand]`), `upper`, `body`:
+
+```python
+ai = Model("teach")                       # or Model("가르치기")
+ai.learn("can", "photos/cans/")           # a folder per thing to recognise
+ai.learn("bottle", "photos/bottles/")
+print(ai.guess("new_photo.jpg"))          # ('can', 0.93)
+print(ai.score("test_photos/"))           # accuracy + what it confuses
+ai.save("recycling")                      # -> Documents/ovkit/recycling.json
+for r in ai.predict(0, stream=True): ...  # webcam over YOUR classes
+```
+
+Korean method names work too: `배우기`, `맞혀봐`, `점수`, `저장`. Collect
+examples with the webcam: `from ovkit.pipelines.teach import collect; collect("가위", 30)`.
+
+## Train your own detector
+
+RT-DETR, trainable, Apache-2.0 end to end — an Ultralytics-style workflow with
+no AGPL code or weights anywhere (`pip install "ovkit[train]"`):
+
+```python
+from ovkit import RTDETR
+
+model = RTDETR("rtdetr-r18")                 # or RTDETR("best.pt") to resume
+model.train(data="data.yaml", epochs=100)    # the YOLO-format labels you already have
+model.val(data="data.yaml")                  # mAP50 / mAP50-95
+model.export(half=True)                      # -> IR + labels.txt
+
+r = model("bus.jpg", conf=0.5)               # ovkit Results: print(r), r.found, r.save()
+```
+
+The exported IR drops straight into `Model("path/to/rtdetr-r18.xml")` — the
+`labels.txt` written next to it means your classes answer by name. CLI:
+`ovkit train --data data.yaml`, `ovkit val`, `ovkit export`.
 `ovkit capabilities` prints the list; **`ovkit gui` opens a window** where you can
 click through them against your webcam or a picture.
 

ovkit-0.2.0/PKG-INFO → ovkit-0.3.0/README.md

@@ -1,48 +1,3 @@
-Metadata-Version: 2.5
-Name: ovkit
-Version: 0.2.0
-Summary: A simple Python inference API for OpenVINO: one import, one Model class, clean Results — with AUTO/NPU, async, and INT8.
-Project-URL: Homepage, https://github.com/leeyunjai82/ovkit
-Project-URL: Documentation, https://leeyunjai82.github.io/ovkit/
-Project-URL: Repository, https://github.com/leeyunjai82/ovkit
-Project-URL: Issues, https://github.com/leeyunjai82/ovkit/issues
-Project-URL: Models, https://huggingface.co/leeyunjai/ovkit-models
-Author: ovkit contributors
-License: Apache-2.0
-License-File: LICENSE
-Keywords: computer-vision,detection,inference,npu,openvino,quantization
-Classifier: Development Status :: 3 - Alpha
-Classifier: License :: OSI Approved :: Apache Software License
-Classifier: Operating System :: OS Independent
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3.11
-Classifier: Programming Language :: Python :: 3.12
-Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
-Requires-Python: >=3.10
-Requires-Dist: huggingface-hub>=0.20
-Requires-Dist: numpy>=1.21
-Requires-Dist: opencv-python-headless>=4.5
-Requires-Dist: openvino>=2024.0
-Requires-Dist: pillow>=9.0
-Requires-Dist: pyyaml>=6.0
-Provides-Extra: all
-Requires-Dist: anomalib; extra == 'all'
-Requires-Dist: nncf; extra == 'all'
-Requires-Dist: openvino-genai; extra == 'all'
-Requires-Dist: optimum-intel; extra == 'all'
-Provides-Extra: anomaly
-Requires-Dist: anomalib; extra == 'anomaly'
-Provides-Extra: dev
-Requires-Dist: black; extra == 'dev'
-Requires-Dist: pytest>=7; extra == 'dev'
-Requires-Dist: ruff; extra == 'dev'
-Provides-Extra: genai
-Requires-Dist: openvino-genai; extra == 'genai'
-Requires-Dist: optimum-intel; extra == 'genai'
-Provides-Extra: quant
-Requires-Dist: nncf; extra == 'quant'
-Description-Content-Type: text/markdown
-
 <div align="center">
 
 <img src="https://raw.githubusercontent.com/leeyunjai82/ovkit/main/docs/_static/logo.svg" width="110" alt="ovkit logo"/>
@@ -138,6 +93,11 @@ camera index).
 | `attention` | `1 person looking at: laptop` | gaze + object detection (ray-cast into the boxes) |
 | `anonymize` | the picture with every face pixelated | face (and plate) detection + redaction |
 | `face_match` | `('yunjai', 0.81)` — who this is | embedding + cosine matching against your gallery |
+| `count` | `pencil 3 · cup 1` (optionally one kind only) | detection + per-kind tally |
+| `posture` | `neck 32° — sit up (6s)` | pose + neck angle **over time** |
+| `exercise` | `squat x 12 (down)` | pose + joint-angle hysteresis (squat, push-up) |
+| `attendance` | `present 24/26` + `roll.csv` | face detection + roster matching |
+| `teach` | your own classes, from example photos | embedding + cosine k-NN (5 modes) |
 
 ```python
 from ovkit import Model, list_pipelines
@@ -149,6 +109,45 @@ Model("face_analyze", attributes=("age_gender",))   # configure what runs
 ```
 
 Aliases: `ocr`, `anpr`, `blur`, `driver`, `describe`, `faces`, `people`, `vehicle`, `tracking`, `reid`.
+
+## Teach your own AI (no GPU)
+
+A Teachable Machine in five lines — an embedding model turns examples into
+vectors, new inputs match the nearest ones. Five modes: `photo` (default),
+`face` (expressions), `hand` (needs `ovkit[hand]`), `upper`, `body`:
+
+```python
+ai = Model("teach")                       # or Model("가르치기")
+ai.learn("can", "photos/cans/")           # a folder per thing to recognise
+ai.learn("bottle", "photos/bottles/")
+print(ai.guess("new_photo.jpg"))          # ('can', 0.93)
+print(ai.score("test_photos/"))           # accuracy + what it confuses
+ai.save("recycling")                      # -> Documents/ovkit/recycling.json
+for r in ai.predict(0, stream=True): ...  # webcam over YOUR classes
+```
+
+Korean method names work too: `배우기`, `맞혀봐`, `점수`, `저장`. Collect
+examples with the webcam: `from ovkit.pipelines.teach import collect; collect("가위", 30)`.
+
+## Train your own detector
+
+RT-DETR, trainable, Apache-2.0 end to end — an Ultralytics-style workflow with
+no AGPL code or weights anywhere (`pip install "ovkit[train]"`):
+
+```python
+from ovkit import RTDETR
+
+model = RTDETR("rtdetr-r18")                 # or RTDETR("best.pt") to resume
+model.train(data="data.yaml", epochs=100)    # the YOLO-format labels you already have
+model.val(data="data.yaml")                  # mAP50 / mAP50-95
+model.export(half=True)                      # -> IR + labels.txt
+
+r = model("bus.jpg", conf=0.5)               # ovkit Results: print(r), r.found, r.save()
+```
+
+The exported IR drops straight into `Model("path/to/rtdetr-r18.xml")` — the
+`labels.txt` written next to it means your classes answer by name. CLI:
+`ovkit train --data data.yaml`, `ovkit val`, `ovkit export`.
 `ovkit capabilities` prints the list; **`ovkit gui` opens a window** where you can
 click through them against your webcam or a picture.
 

{ovkit-0.2.0 → ovkit-0.3.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "ovkit"
-version = "0.2.0"
+version = "0.3.0"
 description = "A simple Python inference API for OpenVINO: one import, one Model class, clean Results — with AUTO/NPU, async, and INT8."
 readme = "README.md"
 requires-python = ">=3.10"
@@ -33,7 +33,9 @@ dependencies = [
 genai = ["openvino-genai", "optimum-intel"]
 anomaly = ["anomalib"]
 quant = ["nncf"]
-all = ["openvino-genai", "optimum-intel", "anomalib", "nncf"]
+train = ["torch>=2.0", "torchvision>=0.15", "scipy", "onnx"]
+hand = ["mediapipe"]
+all = ["openvino-genai", "optimum-intel", "anomalib", "nncf", "torch>=2.0", "torchvision>=0.15", "scipy", "onnx"]
 dev = ["pytest>=7", "ruff", "black"]
 
 [project.urls]

{ovkit-0.2.0 → ovkit-0.3.0}/src/ovkit/__init__.py

@@ -40,10 +40,20 @@ from .core.registry import list_models
 from .core.results import Boxes, Keypoints, Masks, Probs, Results
 from .pipelines import Pipeline, list_pipelines
 
-__version__ = "0.2.0"
+__version__ = "0.3.0"
+
+
+def __getattr__(name: str):  # lazy: keeps `import ovkit` free of the train stack
+    if name == "RTDETR":
+        from .rtdetr import RTDETR
+
+        return RTDETR
+    raise AttributeError(f"module 'ovkit' has no attribute {name!r}")
+
 
 __all__ = [
     "Model",
+    "RTDETR",
     "Pipeline",
     "list_pipelines",
     "Results",

{ovkit-0.2.0 → ovkit-0.3.0}/src/ovkit/__main__.py

@@ -1,4 +1,4 @@
-"""``ovkit`` command-line interface: ``gui``, ``run``, ``list``, ``info``, ``download``, ``devices``."""
+"""``ovkit`` CLI: ``gui``, ``run``, ``train``, ``val``, ``export``, ``list``, ``info``, ``download``, ``devices``."""
 
 from __future__ import annotations
 
@@ -132,6 +132,33 @@ def _cmd_gui(args: argparse.Namespace) -> int:
     return gui_main(device=args.device, camera=args.camera)
 
 
+def _cmd_train(args: argparse.Namespace) -> int:
+    """Train RT-DETR on YOLO-format data: ``ovkit train --data data.yaml``."""
+    from .rtdetr import RTDETR
+
+    model = RTDETR(args.model, device=args.device)
+    best = model.train(
+        data=args.data, epochs=args.epochs, imgsz=args.imgsz, batch=args.batch, lr=args.lr
+    )
+    print(f"best checkpoint: {best}")
+    return 0
+
+
+def _cmd_val(args: argparse.Namespace) -> int:
+    from .rtdetr import RTDETR
+
+    RTDETR(args.model, device=args.device).val(data=args.data, imgsz=args.imgsz)
+    return 0
+
+
+def _cmd_export(args: argparse.Namespace) -> int:
+    from .rtdetr import RTDETR
+
+    out = RTDETR(args.model).export(imgsz=args.imgsz, half=args.half, out_dir=args.out)
+    print(f"exported: {out} (+ labels.txt)")
+    return 0
+
+
 def main(argv: list[str] | None = None) -> int:
     """CLI entry point. Returns a process exit code."""
     parser = argparse.ArgumentParser(prog="ovkit", description="ovkit model utilities")
@@ -158,6 +185,30 @@ def main(argv: list[str] | None = None) -> int:
     p_dl.add_argument("--no-convert", action="store_true", help="skip IR conversion")
     p_dl.set_defaults(func=_cmd_download)
 
+    p_train = sub.add_parser("train", help="train RT-DETR on your own data (needs ovkit[train])")
+    p_train.add_argument("--data", required=True, help="data.yaml (YOLO format)")
+    p_train.add_argument("--model", default="rtdetr-r18", help="variant or a .pt to resume")
+    p_train.add_argument("--epochs", type=int, default=100)
+    p_train.add_argument("--imgsz", type=int, default=640)
+    p_train.add_argument("--batch", type=int, default=8)
+    p_train.add_argument("--lr", type=float, default=1e-4)
+    p_train.add_argument("--device", default=None)
+    p_train.set_defaults(func=_cmd_train)
+
+    p_val = sub.add_parser("val", help="mAP on the val split of a data.yaml")
+    p_val.add_argument("--model", required=True, help="a trained .pt")
+    p_val.add_argument("--data", required=True)
+    p_val.add_argument("--imgsz", type=int, default=640)
+    p_val.add_argument("--device", default=None)
+    p_val.set_defaults(func=_cmd_val)
+
+    p_exp = sub.add_parser("export", help="export a trained .pt to OpenVINO IR + labels.txt")
+    p_exp.add_argument("--model", required=True, help="a trained .pt")
+    p_exp.add_argument("--imgsz", type=int, default=640)
+    p_exp.add_argument("--half", action="store_true", help="FP16 IR")
+    p_exp.add_argument("--out", default=".", help="output directory")
+    p_exp.set_defaults(func=_cmd_export)
+
     p_dev = sub.add_parser("devices", help="list OpenVINO devices")
     p_dev.set_defaults(func=_cmd_devices)
 

{ovkit-0.2.0 → ovkit-0.3.0}/src/ovkit/core/i18n.py

@@ -46,6 +46,14 @@ KO_CAPS: dict[str, str] = {
     "얼굴가리기": "anonymize",
     "모자이크": "anonymize",
     "누구지": "face_match",
+    "가르치기": "teach",
+    "내가가르치기": "teach",
+    "개수세기": "count",
+    "거북목": "posture",
+    "거북목알림": "posture",
+    "운동횟수": "exercise",
+    "출석체크": "attendance",
+    "출석": "attendance",
     # single-model aliases (registry)
     "물체찾기": "detect",
     "이건뭐야": "classify",

{ovkit-0.2.0 → ovkit-0.3.0}/src/ovkit/core/results.py

@@ -429,7 +429,7 @@ class Results:
         if self.masks is not None and len(self.masks):
             parts.append(self._mask_summary(max_items))
 
-        if self.keypoints is not None:
+        if self.keypoints is not None and not self.text:
             n, k = self.keypoints.data.shape[0], self.keypoints.data.shape[1]
             parts.append(f"{n} instance(s), {k} keypoints")
 

{ovkit-0.2.0 → ovkit-0.3.0}/src/ovkit/pipelines/__init__.py

@@ -25,11 +25,13 @@ from typing import Any
 from .analyze import FaceAnalyzer, PersonAnalyzer, VehicleAnalyzer
 from .attention import AttentionAnalyzer
 from .base import Pipeline
+from .classroom import Attendance, Counter, PostureCoach, RepCounter
 from .gaze import GazeEstimator
 from .plates import PlateReader
 from .privacy import Anonymizer
 from .reid import ReID
 from .scene import SceneReport
+from .teach import Teach
 from .temporal import DrowsinessMonitor, GestureRecognizer
 from .text import TextReader
 from .tracking import Tracker
@@ -54,6 +56,13 @@ PIPELINES: dict[str, type[Pipeline]] = {
     # identity: match it, or remove it
     "face_match": ReID,
     "anonymize": Anonymizer,
+    # make your own
+    "teach": Teach,
+    # classroom
+    "count": Counter,
+    "posture": PostureCoach,
+    "exercise": RepCounter,
+    "attendance": Attendance,
 }
 
 #: Friendlier spellings people reach for first.
@@ -162,4 +171,9 @@ __all__ = [
     "GestureRecognizer",
     "PlateReader",
     "SceneReport",
+    "Teach",
+    "Attendance",
+    "Counter",
+    "PostureCoach",
+    "RepCounter",
 ]