This repository contains two projects: a Text-to-Speech (TTS) project using Microsoft's SpeechT5 model and a YOLO Object Detector project using the YOLOv5 model.
This project demonstrates the use of the Microsoft SpeechT5 model for text-to-speech synthesis. It provides examples of generating speech from text using pre-trained models and speaker embeddings.
torch
transformers
datasets
soundfile
Install the required packages using:
pip install torch transformers datasets soundfile
-
Initialize the TTS pipeline:
- Use the
pipeline
function fromtransformers
to create a text-to-speech pipeline with themicrosoft/speecht5_tts
model.
- Use the
-
Load speaker embeddings:
- Load speaker embeddings from the
Matthijs/cmu-arctic-xvectors
dataset.
- Load speaker embeddings from the
-
Generate and save speech:
- Generate speech from text and save it as an audio file.
- Initialize the pipeline and synthesizer.
- Load the embeddings dataset and extract a specific embedding.
- Generate speech and save it to a file.
This project is licensed under the MIT License.
This project demonstrates the implementation of the YOLOv5 object detection model. It provides examples of loading the model, preprocessing images, performing object detection, and visualizing the results.
torch
opencv-python
matplotlib
yolov5
(from the official YOLOv5 repository)
Install the required packages using:
pip install torch opencv-python matplotlib git+https://github.com/ultralytics/yolov5.git
-
Load the YOLOv5 model:
- Load the pre-trained YOLOv5 model using the
torch.hub.load
method.
- Load the pre-trained YOLOv5 model using the
-
Preprocess images:
- Preprocess input images to the required format for YOLOv5.
-
Perform object detection:
- Use the model to detect objects in the preprocessed images.
-
Draw bounding boxes and visualize results:
- Draw bounding boxes around detected objects and display/save the result.
- Load and preprocess an image.
- Perform object detection and apply non-max suppression.
- Draw bounding boxes on detected objects and display/save the resulting image.
- George Youhana - [email protected]
- Mostafa Magdy - [email protected]
- Abdallah Alkhouly - [email protected]