This model was created to predict the depth of a single RGB image of food after transfering a depth model that had been used to predict the depths of indoor scenes, to predict the depth of food images.
Our idea is to use our small dataset to create a model that predicts the depth of a food image starting from the pretrained Tensorflow weights.
Some codes of this repository are already implemented in this repositoy
To start training, please follow the following steps:
- open the depth_train_code.ipynb on Google colab
- run all the cells till the cell that starts training, these cells should import the required libraries, clone this repository and download the required model and dataset.
- run the cell that starts training the model.
- run the cell that starts testing the model.
- the output images of training and testing should appear in the same directory that contains this cloned repository
Our training & testing results are in the excel sheet Depth_training_results.xlsx
Also some of the output images of train and tests appear in the folder Training out images
