https://github.com/ksachdeva/face-embeddings-tsne
Face embeddings visualization using t-SNE
https://github.com/ksachdeva/face-embeddings-tsne
computer-vision dlib embeddings face-recognition neural-networks t-sne tensorflow tfjs
Last synced: 8 months ago
JSON representation
Face embeddings visualization using t-SNE
- Host: GitHub
- URL: https://github.com/ksachdeva/face-embeddings-tsne
- Owner: ksachdeva
- License: apache-2.0
- Created: 2018-07-08T21:34:02.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2018-07-10T14:43:53.000Z (about 7 years ago)
- Last Synced: 2025-01-04T08:35:47.095Z (9 months ago)
- Topics: computer-vision, dlib, embeddings, face-recognition, neural-networks, t-sne, tensorflow, tfjs
- Language: TypeScript
- Homepage: https://ksachdeva.github.io/face-embeddings-tsne/
- Size: 1.15 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Face embeddings visualization using t-SNE
** WORK IN PROGRESS **
This is a very simple web page (WIP) that uses the [tfjs-tsne](https://github.com/tensorflow/tfjs-tsne) library to visualize the face embeddings generated using [dlib](https://github.com/davisking/dlib).
See the demo at [https://ksachdeva.github.io/face-embeddings-tsne/](https://ksachdeva.github.io/face-embeddings-tsne/). Make sure to open the javascript console as well to see some debug information. On my macbook pro 2015, if I set perplexity more than 18 I get an error. On a Ubuntu machine with NVIDIA 1080 I am able to set the max perplexity (30) with out any error.
## Setup / Install / Run instructions
A sample file containing the embeddings [dlib.json](embeddings/dlib.json) is already included in this repository. If you want to generate a sample that contains more classes then you can use [face-embeddings-generator](https://github.com/ksachdeva/face-embeddings-generator) project.
This sample file contains 10 classes only.
If you are generating the new sample file then make sure that your file is generated in the *visualizer/src/assets* directory of this project as this is where the javascript expects it to be.
```bash
# if you want to get face-embeddings-generator as part of this project
# then clone it recursively
git clone --recursive https://github.com/ksachdeva/face-embeddings-tsne
```
```bash
# to run
cd visualizer
npm install
ng serve
```