ObjDet Documentation¶
Production-grade object detection training framework built on PyTorch Lightning.
Getting Started
User Guide
API Reference
Development
Features¶
Multiple Models: Faster R-CNN, RetinaNet, YOLOv8/v11
Flexible Data: COCO, VOC, YOLO formats with LitData optimization
Training: LightningCLI, custom callbacks, Optuna tuning
Deployment: LitServe REST API with A/B testing
MLOps: Celery job queue, job dependencies, resource routing
Web UI: React + FastAPI interface for training management
Quick Example¶
from objdet.models import FasterRCNN
from objdet.data import COCODataModule
# Create model
model = FasterRCNN(num_classes=80)
# Create datamodule
datamodule = COCODataModule(
data_dir="/data/coco",
batch_size=16,
)
# Train with Lightning
trainer = Trainer(max_epochs=50)
trainer.fit(model, datamodule)