Evolving-AI-Lab/synthesizing
Research code that synthesizes optimal input images to visualize what specific neurons in a neural network respond to.
Not currently ranked — collecting fresh signals.
star history
This repository implements the technique of synthesizing preferred inputs for neurons in neural networks using deep generator networks (Upconvolutional networks). The approach inverses a neural network to find inputs that maximally activate specific neurons, enabling visualization and understanding of what visual features individual neurons detect. Built on Caffe with Python bindings, it includes downloading pre-trained generator networks and visualization target models.
Frequently asked
- What is Evolving-AI-Lab/synthesizing?
- Research code that synthesizes optimal input images to visualize what specific neurons in a neural network respond to.
- Is synthesizing open source?
- Yes — Evolving-AI-Lab/synthesizing is open source, released under the MIT license.
- What language is synthesizing written in?
- Evolving-AI-Lab/synthesizing is primarily written in Python.
- How popular is synthesizing?
- Evolving-AI-Lab/synthesizing has 471 stars on GitHub.
- Where can I find synthesizing?
- Evolving-AI-Lab/synthesizing is on GitHub at https://github.com/Evolving-AI-Lab/synthesizing.