DeepDream using Tensorflow 2.3
usage: deepdream.py [-h] --input_image INPUT_IMAGE [--output_dir OUTPUT_DIR] [--layer_names LAYER_NAMES [LAYER_NAMES ...]] [--lr LR] [--octave_scale OCTAVE_SCALE] [--num_octaves NUM_OCTAVES] [--iterations ITERATIONS] optional arguments: -h, --help show this help message and exit --input_image INPUT_IMAGE Path to an input image --output_dir OUTPUT_DIR Path to a directory where the output will be saved. Default: Same directory in 'outputs' dir. --layer_names LAYER_NAMES [LAYER_NAMES ...] Layer at which we modify image to maximize outputs. Choose either: 'mixed0', 'mixed1', ...,'mixed10' or any combination of them. Default: ['mixed3', 'mixed5'] --lr LR Learning rate Default: 0.01 --octave_scale OCTAVE_SCALE Image scale between octaves Default: 0.75 --num_octaves NUM_OCTAVES Number of octaves; How many downsampling to do and apply DeepDream. Default determines the maximal num octaves possible. --iterations ITERATIONS Number of Gradient Ascent steps per octave. Default: 10
Checkout the file names for what intermediate layers were used.