Skip to content

Demo

1. Asynchronous Inference Server with FastAPI

First, run the backend inference server

$python inference_server.py
INFO:     Started server process [18987]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)

Then, on the different process, run the front end web app via streamlit. Specify the address of the inference server you booted up above by -s option. For example, if the server is booted at http://127.0.0.1:8000, pass 8000.

$streamlit run demo/webapp_demo.py -- -s  8000

2. Realtime Inference

This requires a webcam connected to your computer.

First, uncomment the mode you want to try and comment all others out. Following example chose MiDAS to try.

model = BiWAKO.MiDAS("weights/mono_depth_small")
# model = BiWAKO.U2Net("mobile")
# model = BiWAKO.HumanParsing("human_attribute")

Now, just run the script

$python demo/live_demo.py

3. Video Inference

Sample usage of VideoPredictor. Assign the path to input mp4 to video and run the script.

$python demo/video_inference_demo.py