Copyright (C) 2019, Zhihao Xia and Ayan Chakrabarti
This distribution provides a Tensorflow implementation, along with trained models, for the method described in our NeurIPS 2019 paper:
Zhihao Xia and Ayan Chakrabarti, "Training Image Estimators without Image Ground-Truth", NeurIPS 2019 (spotlight).
If you find the code useful for your research, we request that you cite the above paper. Please contact [email protected] with any questions.
We evaluate our method on two applications: compressive sensing and face deblurring. The code and pre-trained models for each application can be found in compressive_sensing/ and face_deblur/, respectively.
- Linux or OSX
- NVIDIA GPU + CUDA CuDNN (CPU mode might work, but untested)
- Python3 & Tensorflow (the code has been tested for Tensorflow 1.7 and 1.14)