fbcnn

FBCNN : Towards Flexible Blind JPEG Artifacts Removal

Input

(from https://live.ece.utexas.edu/research/quality/)

Ailia input shape: (1, 3, 256, 256)
Range: [0, 1]

Output

Model 1: color(default)

Model 2: color_real

Model 3: gray

Model 4: gray_doublejpeg

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

python3 fbcnn.py

In the above, model is color, input image is input.bmp and output image is output.bmp.

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.
Furthermore, you can use --model option to change the model and choose model name from [color, color_real, gray, gray_doublejpeg].

$ python3 fbcnn.py --model MODEL_NAME --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --video option, you can input the video. If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 fbcnn.py --model MODEL_NAME --video VIDEO_PATH --savepath SAVE_VIDEO_PATH

Reference

Paper: Towards Flexible Blind JPEG Artifacts Removal (FBCNN, ICCV 2021)
Github: jiaxi-jiang/FBCNN
Dataset: Laboratory for Image & Video Engineering

Framework

PyTorch 1.7.1

Model Format

ONNX opset = 10

Netron

fbcnn_color.onnx.prototxt
fbcnn_color_real.onnx.prototxt
fbcnn_gray.onnx.prototxt
fbcnn_gray_doublejpeg.onnx.prototxt