Project Name: Virtual Object Detector
The Metaverse will have a great amount of AIs, currently AIs will need metadata from the world to detect objects around them, although using a closed-source metaverse might be hard to add metadata if it doesn’t include for your agent, making the agent unable to look around. The Virtual Object Detector can help with this, it doesn’t need to be mount in anything, it accepts any type of images, detects objects inside and their position on the screen if needed, while the suggested use right now is in the Computer, it can also be used in the real world with real-time video recordings.
Virtual Object Detector is an AI (agent) assistant for Virtual (but can be used for physical AIs too) detect objects in real time.
It is using python and ImageAI, there is the option to train the AI with custom targeted images, or use some ready models.
It is using screenshots or images and the tools to capture these images, also it can detect the current point of the images in the screen from a screenshot, capture images from the web using puppeteer.
About the trainer:
The trainer works simply just by making folders with the name of the target and just images inside, no need for CSV or hardcoded data.
Examples of usage:
Can be used for Web3 AI Agents (chat bots etc) be able to detect objects around them without the need of metadata
Real world object detection for robots
Can be used in games for analytics in tournaments
- Open Source Documentation
- Easy Model Trainer Tool - web tool
- Support for Image Capturer from Puppeteer, Full Screen, Games, Video To Images decoder
- More can be added based on people’s ideas
Repository: GitHub - alextitonis/Cognitive-Agent 1
Current state is a little simple, just has the object detector and some maths utils
Some Tests: https://imgur.com/a/0vgiGgJ