Smooth tiled predictions

Do smooth predictions on an image from tiled prediction patches.

biapy.data.post_processing.smooth_tiled_predictions.create_gen(subdivs, subdivs_m, subdivs_w)[source]

biapy.data.post_processing.smooth_tiled_predictions.predict_img_with_smooth_windowing(input_img, window_size, subdivisions, n_classes, pred_func)[source]: Apply the pred_func function to square patches of the image, and overlap the predictions to merge them smoothly. See 6th, 7th and 8th idea here: http://blog.kaggle.com/2017/05/09/dstl-satellite-imagery-competition-3rd-place-winners-interview-vladimir-sergey/

biapy.data.post_processing.smooth_tiled_predictions.predict_img_with_overlap(input_img, window_size, subdivisions, n_classes, pred_func)[source]: Based on predict_img_with_smooth_windowing but works just with the original image instead of creating 8 new ones.