DepthHist Model

Overview

This repository contains the DepthHist model, a preliminary version designed for depth estimation using a histogram-based approach. The model offers two configurations:

• Simple Mode: A basic encoder-decoder architecture.
• Non-Simple Mode: An advanced model that includes a histogram layer for enhanced depth computation.

Features

• Encoder-Decoder Architecture: Leverages backbone models like EfficientNet or ResNet for feature extraction.
• Histogram Layer: (Non-Simple Mode) Calculates depth by processing the encoder-decoder output through a histogram layer.
• Flexible Backbone Support: Easily switch between different backbone architectures.

Installation

To get started, clone the repository and install the required dependencies:

git clone https://github.com/chaouki-ai/DepthHistV01
pip install -r requirements.txt

Usage

Here is a basic example of how to use the DepthHist model:

import torch
from Models.model import DepthHist

#  Example usage
model = DepthHist.build(bins=10, simple=False, backbone="efficientnet")
input_tensor = torch.randn(1, 3, 256, 256)  # Example input
output = model(input_tensor)

print(output.shape)  # Output depth map shape

Contact

This is a primer version of a model used for depth estimation using histogram approaches. For questions, suggestions, or collaborations, feel free to contact me:

• Email: medchaoukiziara@gmail.com
• Email: me@chaouki.pro

Contributing

If you'd like to contribute to this project, please fork the repository and submit a pull request. Contributions are welcome!

License

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