Demo

Our method can tell the difference between falling down and crouhing.

Requirements

pytorch >= 1.2.0

cuda10.0

cudnn7.6.5

Pretrained weights

1.LiteFlowNet

2.YOLOV3

Run

Cd/Your Code PATH
python detect.py --source "Your image or video PATH" --weights "YOLOv3 Weights PATH" --LiteFlowNet_weights "Your FlowNet weights" --view-img "True or False"

Citation

[1] @misc{pytorch-liteflownet,
                  author = {Simon Niklaus},
                  title = {A Reimplementation of {LiteFlowNet} using {Pytorch}},
                  year = {2019},
                  howpublished = {\url{https://github.com/xiaobin1231/Fall-Detection-By-YOLOV3-and-LiteFlowNet}}

[2] @misc{yolov3,
                  author = {ultralytics},
                  title = {YOLOv3 in PyTorch>ONNX>CoreML>iOS},
                  year = {2019},
                  howpublished = {\url{https://github.com/xiaobin1231/Fall-Detection-By-YOLOV3-and-LiteFlowNet}}