Tao Train Ocdnet

Wraps NVIDIA's TAO OCDNet model for arbitrary-oriented text detection in natural scenes using differentiable binarization. The skill handles train, evaluate, inference, export, prune, and quantize workflows with full AutoML support enabled by default. Data prep requires extracting archives into split folders with img/ and gt/ subdirectories, and every action needs explicit dataset path overrides since nothing unpacks at runtime. Training monitors hmean and outputs Lightning checkpoints you can route to pruning or TensorRT deployment. The AutoML integration is solid but requires disabling warmup epochs for single-epoch smoke runs to avoid trainer failures. If you need rotated bounding boxes for storefront signs or document photos, this does the job without making you build the pipeline from scratch.

skills.sh installs: 1.6K
GitHub stars: 3.9K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add nvidia/skills --skill tao-train-ocdnet --agent claude-code

Installs into .claude/skills of the current project.

Files

OCDNet

Standalone install? If this session was not initialized by the TAO skill bank plugin, run the tao-setup skill first (host preflight, credentials, cross-skill discovery).

OCDNet for scene text detection. Detects arbitrary-oriented text regions in natural images using a differentiable binarization approach.

Set model.pretrained_model_path for pretrained weights.

For TAO Deploy TensorRT actions (gen_trt_engine, TensorRT evaluate, and TensorRT inference), read references/tao-deploy-ocdnet.md first. Deploy spec templates live in this skill's references/ folder with the spec_template_deploy_*.yaml prefix.

The PyT OCDNet CLI supports train, evaluate, export, inference, prune, quantize, and default_specs. It does not expose PyT-side retrain or gen_trt_engine subcommands. The model skill exposes retrain by running ocdnet train with model.load_pruned_graph: true and model.pruned_graph_path. Resume from an epoch checkpoint uses ocdnet train plus train.resume_training_checkpoint_path. TensorRT engine generation is owned by the deploy workflow.

Dataclass Schemas

SKILL.md on GitHub

Details

Repository: github.com/nvidia/skills
First seen: Jul 14, 2026