VIDEO LINK: https://drive.google.com/file/d/1VDUq3U8dLKs6ZhImQUvlDP7UO9prbjXq/view?usp=sharing
DATASET: Dataset is taken from the DIODE (Dense Indoor and Outdoor DEpth) dataset, Since training data is too large run the train.py file to download the training and validation dataset
Create a new virtual env (pip)
python -m venv <virtual_env>
cd Scripts\activate
Create a new virtual env (conda)
conda create -n <virtual_env> python=3.8 anaconda
conda activate <virtual_env>
Install dependencies
pip install -r requirements.txt
The model used to train the dataset is from U-Net architecture
Run the train.py file and tune the hyperparameters to train your own model or use the existing parameters to train the U Net model
Use the test.py file and provide the trained model as the parameter to visualize depth maps and compare with the original images Use the visualize_single_image function to visualize single depth maps or use visulaize_depth_map to pass images in a batch