search

iwa-shi / fidelity_controllable_compression

Official PyTorch Implementation of Fidelity Controllable Extreme Image Compression with GAN
MIT License
30 stars 6 forks source link

pre-training model #2

Open studentliuyx opened 2 years ago

studentliuyx commented 2 years ago

Hello, may I ask how to use the pre-trained model you provided。 Thank you in advance!

studentliuyx commented 2 years ago

AND I want to know Why are some of the parameters in the code different from the ones in your pape，such as lamb_mse2，lamb_adv，lamb_vgg. In the code, they are 0.5 0.05 1..but in the paper, they are 0.01, 1, 20.... Thanks!

xinlizhao commented 2 years ago

I tested with the pre training model, but found that the test results could not reconstruct the image at all. Can you tell me what this is all about? I operate in strict accordance with the given tutorials

xinlizhao commented 2 years ago

I tested with the pre training model, but found that the test results could not reconstruct the image at all. Have you reproduced the author's experimental results

studentliuyx commented 2 years ago

I tested with the pre training model, but found that the test results could not reconstruct the image at all. Have you reproduced the author's experimental results

There is something wrong with the code of this article. I have tried many times, but all failed, which wasted a lot of my time. I have given up his code, and I suggest you not to use it. My personal feeling is that there is a problem with compress.py and decompress.py.

studentliuyx commented 2 years ago

I tested with the pre training model, but found that the test results could not reconstruct the image at all. Can you tell me what this is all about? I operate in strict accordance with the given tutorials

I also tried to train the model from scratch for 7 days and still failed. It was really infuriating.

xinlizhao commented 2 years ago

Well, thank you for replying to me. Let's see if the author will change the code in the future. By the way, are you also a student in the direction of generating compression? We can talk about it together

iwa-shi commented 2 years ago

I’m sorry for the late reply. I'm fixing the source code (compress.py and decompress.py). I would appreciate if you could wait for a few days.

xinlizhao commented 2 years ago

Thank you for your reply. I am very interested in your project. I hope you can update it as soon as possible. I will pay attention to it. Thank you very much!!

iwa-shi commented 1 year ago

Hi. I fixed our source code. Please try the command below.

    pip install range-coder
    python codec.py encode -c CONFIG_PATH -m MODEL_PATH -i IMAGE_PATH -o BIN_PATH
    python codec.py decode -c CONFIG_PATH -m MODEL_PATH -i BIN_PATH -o RECONSTRUCTION_PATH

If you still have problems, please let me know. Thank you!