Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

A quick example to set up a monitor on a live stream #127

Merged
merged 4 commits into from
Oct 24, 2023
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
74 changes: 74 additions & 0 deletions docs/docs/getting-started/5-streaming.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,74 @@
# A Quick Example: Live Stream Alert

A quick example to get used to setting up detectors and asking good questions: set up a monitor on a live stream.

## Requirements

- [Groundlight SDK](/docs/installation/) with Python 3.7 or higher
- The video ID of a YouTube live stream you'd like to monitor

## Installation

Ensure you have Python 3.7 or higher installed, and then install the Groundlight SDK and OpenCV library:

```bash
pip install groundlight pillow ffmpeg yt-dlp typer
```

## Creating the Application

1. Save this command as a shell script `get_latest_frame.sh`:

```
#!/bin/bash

ffmpeg -i "$(yt-dlp -g $1 | head -n 1)" -vframes 1 last.jpg -y
```

This will download the most recent frame from a YouTube live stream and save it to a local file `last.jpg`.

2. Log in to the [Groundlight application](https://app.groundlight.ai) and get an [API Token](api-tokens).

3. Next, we'll write the Python script for the application.

```python notest
import os
import subprocess
import typer
from groundlight import Groundlight
from PIL import Image


def main(*, video_id: str = None, detector_name: str = None, query: str = None, confidence: float = 0.75, wait: int = 60):
"""
Run the script to get the stream's last frame as a subprocess, and submit result as an image query to a Groundlight detector
:param video_id: Video ID of the YouTube live stream (the URLs have the form https://www.youtube.com/watch?v=<VIDEO_ID>)
:param detector_name: Name for your Groundlight detector
:param query: Question you want to ask of the stream (we will alert on the answer of NO)
"""
gl = Groundlight()
detector = gl.create_detector(name=detector_name, query=query, confidence_threshold=confidence)

while True:
p = subprocess.run(["./get_latest_frame.sh", video_id])
if p.returncode != 0:
raise RuntimeError(f"Could not get image from video ID: {video_id}. Process exited with return code {p.returncode}.")

image = Image.open("last.jpg").convert("RGB")
response = gl.submit_image_query(detector=detector, image=image, wait=wait)

if response.result.label == "NO":
os.system("say 'Alert!'") # this may not work on all operating systems


if __name__ == "__main__":
typer.run(main)

```

4. Save the script as `streaming_alert.py` in the same directory as `get_latest_frame.sh` above and run it:

```bash
python streaming_alert.py <VIDEO_ID> --detector_name <DETECTOR_NAME> --query <QUERY IN QUOTATION MARKS>
```