Furniture Image Classification

A web-based application that allows users to classify furniture images using multiple pre-trained models. The app offers the flexibility to upload your own image of any extension and size or choose from a set of sample images. It is powered by Streamlit and is designed to showcase the capabilities of deep learning in image classification.

Features

1. Multiple Model Options:

   • Users can select from multiple trained models to classify furniture images.
   • Models include popular architectures trained on custom furniture datasets (Densenet201,VGG16, Resnet50).

2. Image Upload:

   • Upload your own furniture image for classification.
   • Supported file formats: .jpg, .png, .jpeg.

3. Sample Images:

   • Choose from a variety of pre-loaded sample images to test the models.

4. Custom Dataset:

   • Models were trained on a custom dataset of furniture images collected and annotated first-hand.
   • Data augmentation techniques were used to enhance the dataset for improved performance.

5. Deployment:

   • Deployed on Streamlit Cloud.
   • Docker Image Available on Docker Hub

How It Works

1. Select a Model:

   • Choose a model from the dropdown menu.

2. Upload or Select an Image:

   • Upload an image file or select a sample image provided in the app.

3. View Prediction:

   • Click the "Predict" button to view the classification result.
   • The predicted furniture class is displayed along with confidence scores.

Models Used

The following models were trained and deployed:

• DenseNet
• ResNet
• VGG

These models were trained on a custom dataset of furniture images and tested for accuracy and reliability.

Dataset

• Collection:
  • Images were collected from various sources and annotated manually.
• Augmentation:
  • Data augmentation techniques such as rotation, flipping, and scaling were applied to increase dataset diversity.
• Training and Testing:
  • The models were trained on a split of the dataset with a rigorous testing phase to ensure high performance.

Installation (For Local Deployment)

1. Clone the repository:

   git clone https://github.com/AnantVerma-58/Furniture-Classification.git
   cd Furniture-Classification

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   streamlit run app.py

Deployment

The app is deployed on Streamlit Cloud and can be accessed here.

Docker Image

This application is also available as a Docker image, making it easy to run locally without setting up the environment manually.

• Docker Hub Repository: Furniture Image Classification
• Pull the Image:

  docker pull anant58/furnitureimage:1.3
  docker run -d -p 9999:9999 anant58/furniture-classification:1.3

• Acess The app a t http://localhost:9999

File Structure

/Furniture-Classification
├── .gitattributes
├── app.py
├── augment.py
├── densenet.pt
├── evaluate.py
├── model.py
├── pred.py
├── prediction.py
├── requirements.txt
├── resnet.pt
├── run.py
├── split.py
├── train.py
├── transform.py
├── trials.ipynb
├── utils.py
├── vgg.pt
|-- README.md             # Project documentation

Requirements

• Python 3.12
• Libraries: torch, torchvision, opencv-python-haedless, streamlit, numpy

Future Improvements

• Add more pre-trained models for classification.
• Expand the dataset to include more furniture categories.
• Implementing real-time prediction using a webcam.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

---