Awesome

Distorted-Image-With-Flow

• Synthesized images have same height and width (i.e., 1024 x 960). Moreover, our ground-truth flow has three channels. For the first two channels, we define the displacement (∆x, ∆y) at pixel-level which indicate how far each pixel have to move to reach its position in the undistorted image as the rectified Ground-truth. For the last channel, we represent the foreground or background by using the categories (1 or 0) at pixel-level.
• Run python perturbed_images_generation_multiProcess.py --path=./scan/new/ --bg_path=./background/ --output_path=./output/

visualization

Origin / Distorted / Distorted

<img src="https://github.com/gwxie/Distorted-Image-With-Flow/blob/main/output/scan/new_0.png" height="300"/> <img src="https://github.com/gwxie/Distorted-Image-With-Flow/blob/main/output/png/new_0_7_curve.png" height="300"/> <img src="https://github.com/gwxie/Distorted-Image-With-Flow/blob/main/output/png/new_0_7_fold.png" height="300"/>

Origin / Distorted / Distorted

<img src="https://github.com/gwxie/Distorted-Image-With-Flow/blob/main/output/scan/new_1.png" height="240"/> <img src="https://github.com/gwxie/Distorted-Image-With-Flow/blob/main/output/png/new_1_7_curve.png" height="240"/> <img src="https://github.com/gwxie/Distorted-Image-With-Flow/blob/main/output/png/new_1_7_fold.png" height="240"/>

OTHER

For more information, please enter url.

Sorry for the late reply.