LookOutAI is a cutting-edge tool designed for "looking out"; both literally and figuratively. It uses advanced image recognition technology to identify a person in a photo or video clip by matching their image. Once identified, the system can provide a detailed description of the individual, including their actions or behavior.
LookOutAI also offers versatile features, such as selectively censoring a target or uncensoring only the target's face, ensuring privacy or clarity where needed.
This tool is especially valuable for security applications, where video footage must be processed quickly and with high precision. It enables law enforcement or security teams to prepare evidence while protecting the privacy of uninvolved individuals by blurring their faces.
- PIL
- cv2
- Difflib
- Gradio
- Docker
- Pixtral AI
- face_recognition
Initially
git clone https://github.com/munas-git/LookOutAI.git
cd LookOutAI
source setup.sh
Subsequently
source LookOutAIenv/Scripts/activate
python app/app.py
NB: You will have to create a .env file and include your MISTRAL_API_KEY if you wish to work with the "Describe (Target)" feature.
The Initially commands above will clone the repo, navigate to the right path, create and activate a virtual environment for the system with all dependencies, and then run the Gradio UI. The Subsequently code assumes you have previously run Initially and have every system requirement setup and you're within the base dir... Have fun!
Snapshots of the System... (Click for demo video).
Footnote: I’ve decided not to implement the video processing feature as I’m now shifting my focus to developing my AI/MLOps skills. Along the way, I discovered Azure Cognitive Services and its impressive ability to handle such tasks effortlessly. This has inspired me to dive deeper into cloud computing and explore how to leverage its full potential.
It’s all about working smarter, not harder... though I promise, that doesn’t mean I’m lazy! It’s simply about advancing alongside technology and making the most of what’s available.