Mars-Crater-Detection

Implementing YOLO to detect craters on Mars.

(Adapted from https://github.com/experiencor/basic-yolo-keras , many thanks to him for his help.)
Pre-trained weights are provided in his repo.

Steps for setting up the training:

0. Download the data:

Run: python download_data.py This will create a directory named data and downloads the images and the labels.

1. Extract the images and create annotations:

Run: python save_images.py --dataset and python annotation.py --dataset
Note that the choice of parameter dataset should be either train or test.

2. Launch the training:

Run: python train.py

I could not get the loss to converge without the pre-trained weights. They are necessary for a good initialization. The training was done on a NVIDIA GTX 1080 Ti. Some sample detections from the images in the test set are shown above.