Using frozen C3D features plus a linear SVM reportedly reaches 97.8% average accuracy on GMDCSA and 93.7% on CAUCAFall, though the GMDCSA metrics are internally contradictory.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Human Fall Detection using Transfer Learning-based 3D CNN
Using frozen C3D features plus a linear SVM reportedly reaches 97.8% average accuracy on GMDCSA and 93.7% on CAUCAFall, though the GMDCSA metrics are internally contradictory.