Update README.md

README.md

@@ -1,26 +1,28 @@
+---
+license: mit
+language:
+- en
+metrics:
+- accuracy
+- f1
+- precision
+base_model:
+- bhadresh-savani/distilbert-base-uncased-emotion
+pipeline_tag: audio-classification
+tags:
+- speech
+- emotion
+- ser
+- classification
+---
 # TuBERT: Multimodal Speech Emotion Recognition For Real-Time Avatar Control
 
-*This project was developed for my senior thesis at Princeton University. You can read the paper here.*
-
-![Screenshot of the TuBERT GUI](/gui.jpg)
+*This project was developed for my senior thesis at Princeton University. Paper being published soon.*
 
 ## About
 TuBERT is a multimodal speech emotion recognition model that runs in real-time and on-device. I designed it with PNGTubers in mind, but there are plenty of other applications for it as well!
 
-This repository contains the code to try a TuBERT model, such as the one I used for my thesis, using a (very) basic GUI. If this project ever gets more attention I will consider adding more features that make the GUI fit for an actual livestream.
-
-This repository also includes the code I used to train ([instructions](./model/setup/README.md)) and test ([instructions](./model/stats/README.md)) the model.
-
-## Installation
-If this project gets more attention I will also consider making a packaged version of TuBERT and the GUI you can install as an application with one click. Installation should be pretty painless either way, through.
-
-1. Clone the repository: `git clone https://github.com/YacoubKahkajian/TuBERT.git`
-2. Open the cloned folder: `cd TuBERT`
-3. Download a pre-trained TuBERT model from HuggingFace and place it in `/model`: 
-4. `python -m venv .venv && source .venv/bin/activate`
-5. `pip install -e .`
+To test the model for yourself using a GUI, see the [GitHub repository](https://github.com/YacoubKahkajian/TuBERT) for installation instructions.
 
 ## Usage
-After you have followed the instructions above to install TuBERT, run the command `python main.py` inside the repo directory to open the TuBERT GUI. 
-
-I have only tested the TuBERT model on Mac and Linux and the TuBERT GUI on Mac. I can't think of a reason TuBERT shouldn't be able to run on Windows, but let me know if there's any compatibility issues you run into regardless.
+`tubert.pt` is the base TuBERT model trained on MELD, described in the paper and used by default for evaluation. `tubert_iemocap.pt` is the version of the model fine-tuned on IEMOCAP.