/Enlighten-GAN_test

improving low-light image using Enlighten-GAN

Primary LanguageJupyter NotebookGNU Lesser General Public License v3.0LGPL-3.0

EnlightenGAN-inference

Very minimalistic wrapper for EnlightenGAN inference. It uses carefully converted pretrained weights (+ baked in preprocessing) from the original repo and only requires onnxruntime as inference engine.

Installation

pip install git+https://github.com/arsenyinfo/EnlightenGAN-inference

Usage

from enlighten_inference import EnlightenOnnxModel
import cv2

img = cv2.imread('/path/to/image.jpg')
model = EnlightenOnnxModel()

processed = model.predict(img)

Test result

  1. src1

  1. dst1

dst1_screenshot_01 08 2021

  1. src2

  1. dst2

dst2_screenshot_01 08 2021

  1. src3

  1. dst3

dst3_screenshot_02 08 2021

  1. src4

  1. dst4

크기변환 dst4_screenshot_02 08 2021

  1. src5

  1. dst5

dst5_screenshot_02 08 2021

  1. src6

  1. dst6

dst6_screenshot_02 08 2021

  1. src7

  1. dst7

dst7_screenshot_02 08 2021

  1. src8

  1. dst8

dst8_screenshot_02 08 2021

  1. src9

  1. dst9

dst9_screenshot_02 08 2021

  1. src10

  1. dst10

dst10_screenshot_03 08 2021

  1. src11

  1. dst11

크기변환 dst11_screenshot_03 08 2021

  1. src12

  1. dst12

dst12_screenshot_11 08 2021

  1. src13

  1. dst13

dst13_screenshot_11 08 2021

  1. src14

  1. dst14

dst14_screenshot_11 08 2021

  1. src15

  1. dst15

dst15_screenshot_11 08 2021