open_clip and cliport
CLIPort builds upon open_clip by using CLIP embeddings as its vision-language foundation for robotic manipulation tasks, making them complements rather than competitors.
About open_clip
mlfoundations/open_clip
An open source implementation of CLIP.
This project provides pre-trained models that understand both images and text, allowing you to connect what you see with descriptive phrases. You can input an image and a list of text descriptions to get back probabilities of which description best matches the image. This is ideal for researchers or developers building applications that need to categorize images based on natural language or search for images using text.
About cliport
cliport/cliport
CLIPort: What and Where Pathways for Robotic Manipulation
This project helps robotics engineers and researchers train robotic arms to perform complex tabletop manipulation tasks, such as stacking blocks or picking up specific objects. You provide demonstrations of the desired actions, and the system learns a language-conditioned policy, allowing the robot to understand and execute commands like "stack the red block on the blue block." The output is a trained robotic agent capable of precise object handling based on human instructions.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work