Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/icoxfog417/kemono_puyo
:cat: Take kemono pictures and lines up 3, then tanoshii
https://github.com/icoxfog417/kemono_puyo
kemono-friends object-detection python tensorflow tensorflow-models
Last synced: 29 days ago
JSON representation
:cat: Take kemono pictures and lines up 3, then tanoshii
- Host: GitHub
- URL: https://github.com/icoxfog417/kemono_puyo
- Owner: icoxfog417
- License: mit
- Created: 2017-09-30T03:11:40.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2017-10-03T08:29:00.000Z (about 7 years ago)
- Last Synced: 2024-09-27T21:20:16.215Z (about 1 month ago)
- Topics: kemono-friends, object-detection, python, tensorflow, tensorflow-models
- Language: JavaScript
- Homepage: https://www.youtube.com/watch?v=TetxYBS316M
- Size: 1.72 MB
- Stars: 94
- Watchers: 4
- Forks: 6
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# kemono puyo
Take kemono pictures and lines up, then tanoshii!
This is a toy application that made at the hackathon. It is inspired by following two concepts.
* [Kemono Friends](https://en.wikipedia.org/wiki/Kemono_Friends)
* The Kemono Friends is the Japanese animation that the story is like innocent animal character's adventure (`Kemono` means animal in Japanese).
* `Tanoshii` means so fun! It is the keyword in the Kemono Friends.
* [Puyo Puyo](https://en.wikipedia.org/wiki/Puyo_Puyo)
* The Puyo Puyo is a puzzle game.
* It needs 4 Puyos to remove these originally, but I use 3 for simplify
At first I tried to implement the feature to control the falling objects but to see the animal faces is so fun and cute. Because of this, I dropped its implementation.
## Architecture
1. Take picture on your mobile phone etc.
2. Then [IFTTT](https://ifttt.com/discover) sends it to `kemono_puyo` server.
3. `kemono_puyo` detect friends in the picture by [`tensorflow-models/object_detection`](https://github.com/tensorflow/models/tree/master/research/object_detection)
4. Crop the friends faces.
5. Render friends by [Matter.js](http://brm.io/matter-js/)
6. Play friends voice by [HTML5 Audio](https://www.w3schools.com/html/html5_audio.asp)
7. If 3 friends met, they go to park. It is TANOSHII!
Technical Elements summary
* [IFTTT](https://ifttt.com/discover)
* [tensorflow-models/object_detection](https://github.com/tensorflow/models/tree/master/research/object_detection)
* [Matter.js](http://brm.io/matter-js/)
* [HTML5 Audio](https://www.w3schools.com/html/html5_audio.asp)
## Setup
1. git clone `tensorflow/models` on root directory, and make environment to setup it according to the [installation guide](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/installation.md).
2. [Download the model weights](http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_coco_11_06_2017.tar.gz) to `model_weights` folder and expand it.
3. [Download the label file](https://github.com/tensorflow/models/blob/master/research/object_detection/data/mscoco_label_map.pbtxt) to the same folder (`model_weights/ssd_mobilenet_v1_coco_11_06_2017`).
4. Execute `python run_server.py`
If you want connect mobile phone and `kemono_puyo` server, you have to use IFTTT. If requires global ip, [`ngrok`](https://ngrok.com/) will help you.
The `kemono_puyo` supports websocket, so anyone can participate in it (No Nokemono). Let's take photo for lonely friends!
