Model Zoo
Pretrained model workflow
Use model zoo configs and checkpoints as a starting point for detection and segmentation experiments.
How to use a model zoo baseline
Choose a task, load a matching configuration, set confidence thresholds and validate predictions on your own sample images before training.
Typical categories
- Faster R-CNN for detection baselines
- Mask R-CNN for instance segmentation
- RetinaNet for one-stage object detection
- Panoptic segmentation models for scene-level outputs
Best practice
Do not treat a pretrained model as production-ready without testing on representative data. Measure false positives, false negatives, latency and hardware requirements.