An implement of google deep dream with tensorflow. Detailed description can be found here
Demo:
- Python3
- OpenCV
- tensorflow 1.0
-python3 main.py --input {input path} --output {output path}
If you don't input any image, it will generate a dream image with noise.
Gradient ascent region has uncertainty, even same image with same parameters can generate different pictures.
Larger "iter_num" means a more surprising and more different image.
larger receptive field means more semantic information.
To different image, best parameters are different.