alex04072000/ObstructionRemoval

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1004 stars · 184 forks observed · 2026-09-03

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

The official TensorFlow implementation of the CVPR 2020 paper 'Learning to See Through Obstructions', which removes obstructions like window reflections, fences, and raindrops from short image sequences captured by a moving camera. It uses motion differences between background and obstruction layers, combining dense optical flow estimation (PWC-Net) with a deep CNN for layer reconstruction.

Use cases

• remove window reflections from a burst of photos
• see through fence occlusions in image sequences
• remove raindrops from photos
• recover background layer behind obstructions using multiple frames
• reproduce CVPR 2020 obstruction removal research results

When to choose

• you need to remove reflections, fences, or raindrops from a short sequence of 5+ images with camera motion
• you want the reference implementation of the CVPR 2020 paper for research
• you can work with TensorFlow 1.x and have a GPU available

When to avoid

• you only have a single image (the method requires a multi-frame sequence)
• you need a maintained production library - this is research code on TensorFlow 1.10 with no license
• you need real-time or mobile performance

Facets

library · maturity maintenance

image-processing computer-vision deep-learning machine-learning computer-vision image-processing deep-learning machine-learning python reflection-removal obstruction-removal optical-flow tensorflow cvpr-2020 research-code image-restoration linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="alex04072000/ObstructionRemoval")