What it does
You (the user) can give the program a URL to a YouTube video, and the program will tell you what's in the video. For example, it'll say that a person was in the video 82.59% of the time and that a car was in the video for 2.22% of the time.
How we built it
The program is entirely in Python. It uses four major packages:
- youtube-dl (https://github.com/ytdl-org/youtube-dl)
- opencv-python (https://github.com/opencv/opencv-python)
- yolov5 (https://github.com/ultralytics/yolov5)
- matplotlib (https://github.com/matplotlib/matplotlib)
The program does this:
- first, youtube-dl downloads the YouTube video
- then, opencv-python extracts out the frames from the video
- then, yolov5 (using from PyTorch Hub) runs inference on the frames/images
- then, matplotlib creates the bar chart
What's the AI/computer vision model running on the images?
YOLOv5 (https://github.com/ultralytics/yolov5) running on the pretrained COCO weights
Run it yourself
Here's the link to the Google Colab: https://colab.research.google.com/drive/1rDhqQo5hCkKmsbRCr6pEZu3EjBKufema?usp=sharing
After signing into your Google account and creating a copy of the Colab notebook, you can then switch out URLs and test the program yourself on YouTube videos that you choose.
Log in or sign up for Devpost to join the conversation.