This repository forked from https://github.com/ultralytics/yolov3 (powerful implementation of yolo (Yoo Only Look Once) algorithm for object detection using tensorflow, keras and torch) for lip detection and it's state. four lip state will detect include laugh, open, close and pucker. the pretrained weights saved in:

models/mine/best.pt

instruction of using pre-trained weights introduced in Tutorial.

Result Example

Requirements

Python 3.8 or later with all requirements.txt dependencies installed, including torch>=1.6. To install run:

$ pip install -r requirements.txt

Tutorial

you can check net using webcam by:

python detect.py --source 0 --weights models/mine/best.pt 

and check for video by:

python detect.py --source path/to/your/video --weights models/mine/best.pt 

some useful additional arguments are:

--output    - path for saving your results(default inference/output/{name})
--img-size  - size of processing images, lower size cause better speed and worse accuracy(default:640)
--conf-thres- object confidence threshold(default=0.25)

-->

Inference

$ python detect.py --source 0  # webcam
                            file.jpg  # image 
                            file.mp4  # video
                            path/  # directory
                            path/*.jpg  # glob
                            rtsp://170.93.143.139/rtplive/470011e600ef003a004ee33696235daa  # rtsp stream
                            rtmp://192.168.1.105/live/test  # rtmp stream
                            http://112.50.243.8/PLTV/88888888/224/3221225900/1.m3u8  # http stream
                   --weights models/mine/best.pt

Citation