telekinesis-iris 0.0.3__tar.gz

telekinesis_iris-0.0.3/PKG-INFO

@@ -0,0 +1,179 @@
+Metadata-Version: 2.4
+Name: telekinesis-iris
+Version: 0.0.3
+Summary: Model training and inference repository
+Author-email: Telekinesis <support@telekinesis.ai>
+License-Expression: LicenseRef-Proprietary
+Project-URL: Homepage, https://telekinesis.ai
+Project-URL: Documentation, https://docs.telekinesis.ai
+Project-URL: Telekinesis Examples Repository, https://github.com/telekinesis-ai/telekinesis-examples
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Operating System :: OS Independent
+Classifier: Intended Audience :: Developers
+Classifier: Topic :: Software Development :: Libraries
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+Requires-Dist: loguru>=0.7.0
+Requires-Dist: opencv-python>=4.7.0.72
+Requires-Dist: pycocotools>=2.0.6
+Requires-Dist: numpy>=1.24.0
+Requires-Dist: pydantic>=2.0
+Requires-Dist: requests>=2.31.0
+Requires-Dist: rerun-sdk>=0.34.0
+Requires-Dist: scipy>=1.10.0
+Requires-Dist: torch>=2.2.0
+Requires-Dist: torch-hungarian==0.1.0rc0
+Requires-Dist: tensorboard>=2.16.0
+Requires-Dist: tqdm>=4.66.0
+Requires-Dist: transformers<6,>=5.0.0
+Requires-Dist: torchvision>=0.17.0
+Requires-Dist: onnx>=1.16.0
+Requires-Dist: onnxscript>=0.1.0
+Requires-Dist: onnxruntime>=1.18.0
+Requires-Dist: torchmetrics[detection]<1.9.0,>=1.8.2
+Provides-Extra: docs
+Requires-Dist: mkdocs<2,>=1.6; extra == "docs"
+Requires-Dist: mkdocstrings[python]<1,>=0.25; extra == "docs"
+Requires-Dist: mkdocs-awesome-pages-plugin<3,>=2; extra == "docs"
+Provides-Extra: dev
+Requires-Dist: ruff==0.15.8; extra == "dev"
+Requires-Dist: pylint==4.0.5; extra == "dev"
+Requires-Dist: pytest>=8.0.0; extra == "dev"
+Requires-Dist: pytest-cov>=5.0.0; extra == "dev"
+
+<div align="center">
+  <p>
+    <a href="https://telekinesis.ai" target="_blank">
+      <img
+        width="100%"
+        src="https://assets.telekinesis.ai/logo/telekinesis-banner.png"
+        alt="Telekinesis"
+      >
+    </a>
+  </p>
+
+  <br>
+
+  [![PyPI version](https://img.shields.io/pypi/v/telekinesis-iris)](https://pypi.org/project/telekinesis-iris/)
+  [![License](https://img.shields.io/pypi/l/telekinesis-iris)](https://pypi.org/project/telekinesis-iris/)
+  [![Python versions](https://img.shields.io/pypi/pyversions/telekinesis-iris)](https://pypi.org/project/telekinesis-iris/)
+</div>
+
+<p align="center">
+  <a href="https://github.com/telekinesis-ai">GitHub</a>
+  &nbsp;•&nbsp;
+  <a href="https://www.linkedin.com/company/telekinesis-ai/">LinkedIn</a>
+  &nbsp;•&nbsp;
+  <a href="https://x.com/telekinesis_ai">X</a>
+  &nbsp;•&nbsp;
+  <a href="https://discord.gg/7NnQ3bQHqm">Discord</a>
+</p>
+
+# Telekinesis Iris
+
+Telekinesis Iris is a computer vision library for creating COCO datasets and
+training, exporting, and deploying object detection and instance segmentation
+models.
+
+It includes:
+
+- COCO dataset generation and loading with bounding boxes and segmentation masks
+- Local RF-DETR detection and segmentation model implementations
+- PyTorch training, validation, checkpointing, and TensorBoard logging
+- ONNX export and ONNX Runtime inference
+- Rerun visualization for datasets and predictions
+
+## Release Model
+
+Telekinesis Iris is currently in active development (pre-1.0). APIs may evolve
+between minor releases. Install the latest package version for the newest
+features and fixes.
+
+## Installation
+
+1. Install [Miniconda](https://docs.conda.io/projects/miniconda/en/latest/).
+
+2. Create a Python 3.11 environment:
+
+   ```bash
+   conda create -n telekinesis-iris python=3.11
+   ```
+
+3. Activate the environment:
+
+   ```bash
+   conda activate telekinesis-iris
+   ```
+
+4. Install PyTorch and TorchVision for CUDA 12.8, then install the package:
+
+   ```bash
+   pip install torch torchvision --index-url https://download.pytorch.org/whl/cu128
+   ```
+
+   ```bash
+   pip install telekinesis-iris
+   ```
+
+The Python package is imported as `telekinesis.iris`, while the package published on PyPI is named `telekinesis-iris`.
+
+## Example
+
+Train an RF-DETR segmentation model on a COCO dataset:
+
+```python
+from telekinesis.iris.dataset import COCODataset, ResizeSample
+from telekinesis.iris.models import (
+    RFDETRSegNanoConfig,
+    SegmentationTrainConfig,
+    build_criterion_from_config,
+    build_model_from_config,
+    load_pretrain_weights,
+)
+from telekinesis.iris.trainer import Trainer
+
+model_config = RFDETRSegNanoConfig(num_classes=3)
+train_config = SegmentationTrainConfig(
+    dataset_dir="dataset/train",
+    output_dir="results/seg-nano",
+    epochs=10,
+    batch_size=4,
+)
+
+model = build_model_from_config(model_config, train_config)
+load_pretrain_weights(model, model_config)
+criterion, _ = build_criterion_from_config(model_config, train_config)
+dataset = COCODataset(
+    "dataset/train",
+    transforms=ResizeSample(model_config.resolution),
+    include_masks=True,
+)
+
+Trainer(
+    model=model,
+    criterion=criterion,
+    dataset=dataset,
+    output_dir="results/seg-nano",
+    epochs=10,
+    batch_size=4,
+    evaluate_masks=True,
+).train()
+```
+
+## Resources
+
+- [Telekinesis Examples](https://github.com/telekinesis-ai/telekinesis-examples)
+- [Telekinesis Documentation](https://docs.telekinesis.ai)
+- [Development guide](DEVELOPMENT.md)
+- [Changelog](CHANGELOG.md)
+
+## Support
+
+For issues and questions:
+
+- Create an [issue](https://github.com/telekinesis-ai/telekinesis-examples/issues).
+- Contact `support@telekinesis.ai`.
+- Join the [Telekinesis Discord](https://discord.com/invite/7NnQ3bQHqm).

telekinesis_iris-0.0.3/README.md

@@ -0,0 +1,133 @@
+<div align="center">
+  <p>
+    <a href="https://telekinesis.ai" target="_blank">
+      <img
+        width="100%"
+        src="https://assets.telekinesis.ai/logo/telekinesis-banner.png"
+        alt="Telekinesis"
+      >
+    </a>
+  </p>
+
+  <br>
+
+  [![PyPI version](https://img.shields.io/pypi/v/telekinesis-iris)](https://pypi.org/project/telekinesis-iris/)
+  [![License](https://img.shields.io/pypi/l/telekinesis-iris)](https://pypi.org/project/telekinesis-iris/)
+  [![Python versions](https://img.shields.io/pypi/pyversions/telekinesis-iris)](https://pypi.org/project/telekinesis-iris/)
+</div>
+
+<p align="center">
+  <a href="https://github.com/telekinesis-ai">GitHub</a>
+  &nbsp;•&nbsp;
+  <a href="https://www.linkedin.com/company/telekinesis-ai/">LinkedIn</a>
+  &nbsp;•&nbsp;
+  <a href="https://x.com/telekinesis_ai">X</a>
+  &nbsp;•&nbsp;
+  <a href="https://discord.gg/7NnQ3bQHqm">Discord</a>
+</p>
+
+# Telekinesis Iris
+
+Telekinesis Iris is a computer vision library for creating COCO datasets and
+training, exporting, and deploying object detection and instance segmentation
+models.
+
+It includes:
+
+- COCO dataset generation and loading with bounding boxes and segmentation masks
+- Local RF-DETR detection and segmentation model implementations
+- PyTorch training, validation, checkpointing, and TensorBoard logging
+- ONNX export and ONNX Runtime inference
+- Rerun visualization for datasets and predictions
+
+## Release Model
+
+Telekinesis Iris is currently in active development (pre-1.0). APIs may evolve
+between minor releases. Install the latest package version for the newest
+features and fixes.
+
+## Installation
+
+1. Install [Miniconda](https://docs.conda.io/projects/miniconda/en/latest/).
+
+2. Create a Python 3.11 environment:
+
+   ```bash
+   conda create -n telekinesis-iris python=3.11
+   ```
+
+3. Activate the environment:
+
+   ```bash
+   conda activate telekinesis-iris
+   ```
+
+4. Install PyTorch and TorchVision for CUDA 12.8, then install the package:
+
+   ```bash
+   pip install torch torchvision --index-url https://download.pytorch.org/whl/cu128
+   ```
+
+   ```bash
+   pip install telekinesis-iris
+   ```
+
+The Python package is imported as `telekinesis.iris`, while the package published on PyPI is named `telekinesis-iris`.
+
+## Example
+
+Train an RF-DETR segmentation model on a COCO dataset:
+
+```python
+from telekinesis.iris.dataset import COCODataset, ResizeSample
+from telekinesis.iris.models import (
+    RFDETRSegNanoConfig,
+    SegmentationTrainConfig,
+    build_criterion_from_config,
+    build_model_from_config,
+    load_pretrain_weights,
+)
+from telekinesis.iris.trainer import Trainer
+
+model_config = RFDETRSegNanoConfig(num_classes=3)
+train_config = SegmentationTrainConfig(
+    dataset_dir="dataset/train",
+    output_dir="results/seg-nano",
+    epochs=10,
+    batch_size=4,
+)
+
+model = build_model_from_config(model_config, train_config)
+load_pretrain_weights(model, model_config)
+criterion, _ = build_criterion_from_config(model_config, train_config)
+dataset = COCODataset(
+    "dataset/train",
+    transforms=ResizeSample(model_config.resolution),
+    include_masks=True,
+)
+
+Trainer(
+    model=model,
+    criterion=criterion,
+    dataset=dataset,
+    output_dir="results/seg-nano",
+    epochs=10,
+    batch_size=4,
+    evaluate_masks=True,
+).train()
+```
+
+## Resources
+
+- [Telekinesis Examples](https://github.com/telekinesis-ai/telekinesis-examples)
+- [Telekinesis Documentation](https://docs.telekinesis.ai)
+- [Development guide](DEVELOPMENT.md)
+- [Changelog](CHANGELOG.md)
+
+## Support
+
+For issues and questions:
+
+- Create an [issue](https://github.com/telekinesis-ai/telekinesis-examples/issues).
+- Contact `support@telekinesis.ai`.
+- Join the [Telekinesis Discord](https://discord.com/invite/7NnQ3bQHqm).

telekinesis_iris-0.0.3/pyproject.toml

@@ -0,0 +1,84 @@
+[build-system]
+requires = [
+    "setuptools",
+]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "telekinesis-iris"
+version = "0.0.3"
+description = "Model training and inference repository"
+readme = "README.md"
+requires-python = ">=3.11"
+license = "LicenseRef-Proprietary"
+license-files = [
+    "LICEN[CS]E*",
+]
+authors = [
+    { name = "Telekinesis", email = "support@telekinesis.ai" },
+]
+classifiers = [
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Operating System :: OS Independent",
+    "Intended Audience :: Developers",
+    "Topic :: Software Development :: Libraries",
+]
+dependencies = [
+    "loguru>=0.7.0",
+    "opencv-python>=4.7.0.72",
+    "pycocotools>=2.0.6",
+    "numpy>=1.24.0",
+    "pydantic>=2.0",
+    "requests>=2.31.0",
+    "rerun-sdk>=0.34.0",
+    "scipy>=1.10.0",
+    "torch>=2.2.0",
+    "torch-hungarian==0.1.0rc0",
+    "tensorboard>=2.16.0",
+    "tqdm>=4.66.0",
+    "transformers>=5.0.0,<6",
+    "torchvision>=0.17.0",
+    "onnx>=1.16.0",
+    "onnxscript>=0.1.0",
+    "onnxruntime>=1.18.0",
+    "torchmetrics[detection]>=1.8.2,<1.9.0",
+]
+
+[project.optional-dependencies]
+docs = [
+    "mkdocs>=1.6,<2",
+    "mkdocstrings[python]>=0.25,<1",
+    "mkdocs-awesome-pages-plugin>=2,<3",
+]
+dev = [
+    "ruff==0.15.8",
+    "pylint==4.0.5",
+    "pytest>=8.0.0",
+    "pytest-cov>=5.0.0",
+]
+
+[project.urls]
+Homepage = "https://telekinesis.ai"
+Documentation = "https://docs.telekinesis.ai"
+"Telekinesis Examples Repository" = "https://github.com/telekinesis-ai/telekinesis-examples"
+
+[tool.ruff]
+line-length = 100
+
+[tool.setuptools.packages.find]
+where = [
+    "src",
+]
+namespaces = true
+
+[tool.pytest.ini_options]
+testpaths = [
+    "tests",
+]
+python_files = [
+    "*_test.py",
+]
+addopts = "-ra"

telekinesis_iris-0.0.3/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

telekinesis_iris-0.0.3/src/telekinesis/iris/__init__.py

@@ -0,0 +1 @@
+"""Computer vision dataset, training, export, and inference tools."""

telekinesis_iris-0.0.3/src/telekinesis/iris/dataset.py

@@ -0,0 +1,199 @@
+"""COCO dataset loading and sample transformation utilities."""
+
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Any
+
+import numpy as np
+import torch
+from torch import Tensor
+from torch.utils.data import Dataset
+from PIL import Image
+from pycocotools.coco import COCO
+from torchvision.transforms import InterpolationMode
+from torchvision.transforms import functional as transform_functional
+
+
+class ResizeSample:
+    """Resize an image and its spatial annotations to a square resolution."""
+
+    def __init__(self, resolution: int) -> None:
+        """Initializes a square resize transform.
+
+        Args:
+            resolution: Output image height and width in pixels.
+
+        """
+        self.resolution = resolution
+
+    def __call__(self, image: Tensor, target: dict[str, Any]) -> tuple[Tensor, dict[str, Any]]:
+        """Resizes an image and its spatial annotations.
+
+        Args:
+            image: Image tensor in ``CHW`` layout.
+            target: Detection target containing boxes and optional masks.
+
+        Returns:
+            The resized image and updated target.
+
+        """
+        source_height, source_width = image.shape[-2:]
+        scale_x = self.resolution / source_width
+        scale_y = self.resolution / source_height
+        target = target.copy()
+
+        image = transform_functional.resize(
+            image,
+            [self.resolution, self.resolution],
+            interpolation=InterpolationMode.BILINEAR,
+            antialias=True,
+        )
+
+        boxes = target.get("boxes")
+        if isinstance(boxes, Tensor):
+            boxes = boxes.clone()
+            boxes[:, 0::2] *= scale_x
+            boxes[:, 1::2] *= scale_y
+            target["boxes"] = boxes
+
+        keypoints = target.get("keypoints")
+        if isinstance(keypoints, Tensor):
+            keypoints = keypoints.clone()
+            keypoints[..., 0] *= scale_x
+            keypoints[..., 1] *= scale_y
+            target["keypoints"] = keypoints
+
+        masks = target.get("masks")
+        if isinstance(masks, Tensor):
+            if masks.shape[0] == 0:
+                target["masks"] = masks.new_empty((0, self.resolution, self.resolution))
+            else:
+                target["masks"] = transform_functional.resize(
+                    masks,
+                    [self.resolution, self.resolution],
+                    interpolation=InterpolationMode.NEAREST,
+                )
+
+        area = target.get("area")
+        if isinstance(area, Tensor):
+            target["area"] = area * scale_x * scale_y
+        target["size"] = torch.tensor([self.resolution, self.resolution])
+        return image, target
+
+
+class COCODataset(Dataset):
+    """Load a COCO detection dataset for model training."""
+
+    def __init__(
+        self,
+        root: str | Path,
+        transforms=None,
+        *,
+        include_masks: bool = False,
+    ) -> None:
+        """Initializes a COCO detection dataset.
+
+        Args:
+            root: Root directory containing images and COCO annotations.
+            transforms: Optional callable applied to each image and target.
+            include_masks: Decode segmentation masks for segmentation models.
+
+        Raises:
+            FileNotFoundError: If the annotations file or image directory is missing.
+
+        """
+        self.root = Path(root)
+        rfdetr_annotations = self.root / "_annotations.coco.json"
+        if rfdetr_annotations.is_file():
+            self.images_dir = self.root
+            self.annotations_path = rfdetr_annotations
+        else:
+            self.images_dir = self.root / "images"
+            self.annotations_path = self.root / "annotations.json"
+
+        if not self.annotations_path.is_file():
+            raise FileNotFoundError(f"COCO annotations not found: {self.annotations_path}")
+        if not self.images_dir.is_dir():
+            raise FileNotFoundError(f"Image directory not found: {self.images_dir}")
+
+        self.coco = COCO(str(self.annotations_path))
+
+        self.image_ids = sorted(self.coco.getImgIds())
+
+        self.transforms = transforms
+        self.include_masks = include_masks
+
+    def __len__(self) -> int:
+        """Returns the number of images in the dataset."""
+        return len(self.image_ids)
+
+    def __getitem__(self, index: int) -> tuple[torch.Tensor, dict]:
+        """Loads an image and its detection target.
+
+        Args:
+            index: Zero-based dataset index.
+
+        Returns:
+            An image tensor and target dictionary.
+
+        Raises:
+            ValueError: If masks are requested but an annotation has no segmentation.
+
+        """
+        image_id = self.image_ids[index]
+
+        image_info = self.coco.loadImgs([image_id])[0]
+        image_path = self.images_dir / image_info["file_name"]
+        image = Image.open(image_path).convert("RGB")
+        image = transform_functional.pil_to_tensor(image).float() / 255.0
+        annotation_ids = self.coco.getAnnIds(imgIds=[image_id])
+        annotations = self.coco.loadAnns(annotation_ids)
+
+        boxes = []
+        labels = []
+        areas = []
+        iscrowd = []
+        masks = []
+
+        for annotation in annotations:
+            x, y, width, height = annotation["bbox"]
+
+            boxes.append([x, y, x + width, y + height])
+
+            labels.append(annotation["category_id"])
+
+            areas.append(annotation["area"])
+
+            iscrowd.append(annotation.get("iscrowd", 0))
+
+            if self.include_masks:
+                if not annotation.get("segmentation"):
+                    raise ValueError(f"Annotation {annotation['id']} has no segmentation mask")
+                masks.append(self.coco.annToMask(annotation))
+
+        boxes = torch.tensor(boxes, dtype=torch.float32).reshape(-1, 4)
+
+        labels = torch.tensor(labels, dtype=torch.int64)
+        areas = torch.tensor(areas, dtype=torch.float32)
+        iscrowd = torch.tensor(iscrowd, dtype=torch.int64)
+
+        target = {
+            "boxes": boxes,
+            "labels": labels,
+            "image_id": torch.tensor(image_id, dtype=torch.int64),
+            "area": areas,
+            "iscrowd": iscrowd,
+        }
+
+        if self.include_masks:
+            target["masks"] = (
+                torch.tensor(np.stack(masks), dtype=torch.uint8)
+                if masks
+                else torch.empty((0, image.shape[-2], image.shape[-1]), dtype=torch.uint8)
+            )
+
+        if self.transforms is not None:
+            image, target = self.transforms(image, target)
+
+        return image, target