GitHub  •  LinkedIn  •  X  •  Discord

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.

2. Create a Python 3.11 environment:

   conda create -n telekinesis-iris python=3.11
   

3. Activate the environment:

   conda activate telekinesis-iris
   

4. Install PyTorch and TorchVision for CUDA 12.8, then install the package:

   pip install torch torchvision --index-url https://download.pytorch.org/whl/cu128
   

   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:

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
• Telekinesis Documentation
• Development guide
• Changelog

Support

For issues and questions:

• Create an issue.
• Contact support@telekinesis.ai.
• Join the Telekinesis Discord.