seba-1511 / cervix.kaggle Goto Github PK

The task is to create a new dataset to perform regression on the center of the cervices.

The best approach is to implement a new dataset class. Upon initialization, first properly parse the center_labels.csv file into a dict so that problematic samples can be avoided. Then, use the ImageFolder dataset to load the images. Upon calling __iter__(), the image should be loaded first (if it has a proper center) and then the label of the image should be retrieve from the center_labels dict.

Ideally, this dataset's constructor takes two arguments: the csv file with labels and any ImageFolder dataset. Doing so, we can easily play with the transforms of the image folder dataset, and split the dataset further.

Let me know if there are any issues or if some of the advices are unclear.

One thing we can try is to downsample images from start, and see how well humans can perform on those small size images. (eg, if we only need 1024x1024px images, then we save loading time as opposed to 4096x4096px)

So the task is to try different image sizes, and tell us what you think is the best ration of size reduction vs blurriness of the image.

While finding the optimal image size, it would also be nifty to have a script that takes a folder of images and creates a copy of all images of this folder but in a different location. See this: https://stackoverflow.com/questions/273946/how-do-i-resize-an-image-using-pil-and-maintain-its-aspect-ratio

Please comment your username when you'll be able to write code and be assigned work.

This scripts takes command line arguments (you can modify utils.py#parse_args() accordingly) including a task and a weight file. It loads the weights, goes through all of the test dataset (not the validation dataset from the task itself, so we need to create a function load_test_data(task)) and outputs a csv file name mytask_submit.csv that can be directly submitted on the kaggle website.

I will soon push a baseline model weights that we can use to debug this script.

Thanks for helping on this.