How zhangaocanada/raddet works
RADDet is a TensorFlow radar object-detection research project. Its principal workflow prepares RAD tensors and labels, trains or loads a detector with a 3D RAD head and a Cartesian-box head, then runs inference or evaluation to produce detections, metrics, and visualizations. The README documents both detection spaces and the RadarResNet/VGG backbone options. Some connections below rely on the SOURCE INDEX or README rather than full source excerpts; unsampled implementation details are not inferred.
Main entry points: model.py RAD detector · inference.py Run inference · batch_data_generator.py Batch data generator · evaluate.py Evaluate detections
Components 19
Entry workflows
- Train RAD model
train.py - Train Cartesian head
train_cart.py - Run inference
inference.py - Evaluate detections
evaluate.py
Data handling
- Batch data generator
dataset/batch_data_generator.py - Dataset and config loader
util/loader.py - Geometry and transforms
util/helper.py
Detection models
- RAD detector
model/model.py - Cartesian detector
model/model_cart.py - RadarResNet backbone
model/backbone_radarResNet.py - VGG backbone option
model/backbone_VGG.py - YOLO detection head
model/head_YOLO.py - Neural network layers
model/layers.py - Detection losses
model/loss_functions.py
Metrics and presentation
- mAP metrics
metrics/mAP.py - Detection visualizations
util/drawer.py
Other components
- Researcher
- Radar data
- Metrics and plots
Connections 33
- Researcher → Train RAD model: trains
README.md - Researcher → Run inference: runs
README.md - Radar data → Batch data generator: provides frames
dataset/batch_data_generator.py - Train RAD model → Dataset and config loader: loads configuration
train.py - Train RAD model → RAD detector: trains
train.py - Train RAD model → Batch data generator: requests batches
train.py - Train Cartesian head → RAD detector: builds backbone
train_cart.py - Train Cartesian head → Cartesian detector: trains
train_cart.py - Train Cartesian head → Batch data generator: requests batches
train_cart.py - Run inference → Dataset and config loader: loads configuration
inference.py - Run inference → RAD detector: runs detector
inference.py - Run inference → Cartesian detector: runs detector
inference.py - Run inference → Batch data generator: loads data
inference.py - Run inference → Detection visualizations: plots detections
inference.py - Evaluate detections → Batch data generator: requests batches
evaluate.py - Evaluate detections → RAD detector: runs detector
evaluate.py - Evaluate detections → Cartesian detector: runs detector
evaluate.py - Evaluate detections → mAP metrics: computes AP
evaluate.py - Evaluate detections → Detection visualizations: plots results
evaluate.py - Batch data generator → Dataset and config loader: reads samples
dataset/batch_data_generator.py - Batch data generator → Geometry and transforms: encodes labels
dataset/batch_data_generator.py - RAD detector → RadarResNet backbone: builds backbone
model/model.py - RAD detector → VGG backbone option: offers backbone
model/model.py - RAD detector → YOLO detection head: builds head
model/model.py - RAD detector → Detection losses: computes loss
model/model.py - RadarResNet backbone → Neural network layers: uses layers
model/backbone_radarResNet.py - VGG backbone option → Neural network layers: uses layers
model/backbone_VGG.py - YOLO detection head → Neural network layers: uses layers
model/head_YOLO.py - Cartesian detector → Neural network layers: uses layers
model/model_cart.py - Cartesian detector → Geometry and transforms: uses geometry
model/model_cart.py - mAP metrics → Geometry and transforms: computes IoU
metrics/mAP.py - Run inference → Metrics and plots: produces
inference.py - Evaluate detections → Metrics and plots: produces
evaluate.py