# marcbelmont/cnn-watermark-removal

Fully convolutional deep neural network to remove transparent overlays from images

Repository: https://github.com/marcbelmont/cnn-watermark-removal
Canonical: https://ross.abutalabs.com/products/cnn-watermark-removal
Language: Python
License Family: other
Topics: tensorflow, densenet, dilatednet, watermark, inpainting, fully-convolutional-network
Last push: 2021-03-29T17:02:59+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3259, "days_push": 1983, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1269, forks 221 (observed 2026-08-28T04:04:11.659203+00:00)

## What it is
A TensorFlow implementation of a fully convolutional neural network that removes transparent watermark overlays from images. It trains on synthetic rectangular masks over Pascal VOC/CIFAR images and predicts the original pixels under the overlay.

## Use cases
- remove transparent watermark overlays from images
- inpaint regions covered by semi-transparent text or logos
- train a CNN to reconstruct occluded image areas
- experiment with dilated convolutions for image restoration
- run inference on a watermarked image with a selection mask

## When to choose
- the watermark is a simple additive transparent overlay similar to the synthetic masks
- you want a small, self-contained TensorFlow codebase to study watermark removal
- you need a pretrained model for quick experiments on overlay removal

## When to avoid
- the watermark is complex, tiled, or very different from simple rectangular overlays
- you need production-quality generalizable inpainting - deep image prior methods are recommended by the author
- you require a maintained project with a license and active development
- the overlay is applied non-additively (e.g., multiplicative blending)

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing, computer-vision
- domain: machine-learning, computer-vision, image-processing
- platform: python, cross-platform
- tags: watermark-removal, inpainting, tensorflow, fully-convolutional-network, dilated-convolution, densenet, image-restoration, linux

## Member repositories
- marcbelmont/cnn-watermark-removal (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:11.659203+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T05:03:32.912856+00:00, confidence not recorded.
  - readme: https://github.com/marcbelmont/cnn-watermark-removal (fetched 2026-08-28T04:04:11.659203+00:00, sha 1322a98b0f02)
- Data as of 2026-08-30T08:39:29.467469+00:00.