Sparse4D

Sparse4D detects and tracks objects in 3D from synchronized, calibrated multi-camera images. This documentation covers data preparation, configuration, training, evaluation, and deployment.

For adaptation to calibrated real-world scenes with 2D supervision, refer to 2D-to-3D Geometric Distillation.

• Sparse4D
  • Data Input for Sparse4D
  • Creating an Experiment Specification File
  • dataset
  • model
  • train
  • evaluate
  • visualize
  • inference
  • export
  • Training the Model
  • 2D-to-3D Geometric Distillation
  • Behavior Changes in 7.3.0
  • Evaluating the Model
  • Running Inference on the Model
  • Exporting the Model
  • Quantization
  • Deploying with TAO Deploy