PyTorch Deep Dream

This is an implementation of Deep Dream in PyTorch.

Example of usage

import timm
import torch

from deepdreamer import DeepDreamer
from utils import open_image


dream = DeepDreamer(model_name="inception_v3", layers_names=[], "cuda:0")

img = open_image("example_img.jpg")

dream.set_layers(["Mixed_5b"])
dream_img1 = dream.run_deep_dream_with_octaves(img)

dream.set_layers(["Mixed_5c"])
dream_img2 = dream.run_deep_dream_simple(img)

Output:

Source image	with Mixed_5b layer of InceptionV3	with Mixed_5c layer of InceptionV3

This code is heavily inspired by this TensorFlow tutorial.