activity-classifier@1.2.0

base-action-v2compiled ✓last training complete · elapsed —2 targets not built
follows the selected node · Dataset

Training

Loss · per epochepoch →
1.551.160.780.400.0211530
Accuracy · validation, per epochepoch →
0.950.780.620.460.2911530

solid · train, dashed · validation

All nodesfollows the selected nodesofa-activity-v2 →
09:02:11datasetresolved mmfi-4class@1.0.0 · 80 recordings
09:02:11datasetlabel grp_action · 4 labels
09:02:14preprocess1024 points · 2.5 s window · via OpenPCDet v0.6
09:02:40preprocess880 windows written
09:02:41trainpretrained model base-action-v2 loaded
—trainepoch and loss lines appear here as the trainer emits them
—evaluatetraining report written
—deploymentbuilt ONNX Runtime · A311Y3