IOPaint: Erase Objects from Photos, Locally | GitHub Repos You Should Know #015
Paint over an unwanted object and let IOPaint fill the area.
Repo: Sanster/IOPaint
Classification: Worth Knowing
We tested IOPaint 1.6.0 with its LaMa model on the CPU, locally. The examples show actual model outputs from real photos: removing a beach umbrella and shadow, then a tower. Inputs, masks and outputs were retained. Unmasked pixels remained unchanged. No manual output retouching.
The mask animation illustrates the actual saved selection. It is an edited demonstration of the local backend, not a recording of the web interface or a processing-speed benchmark. AI estimates replacement pixels; inspect the textures and edges. The photo edits are demonstrations, not documentary evidence of the original scenes.
The IOPaint repository is archived. Check compatibility before installing. Other supported models have separate terms.
Licences: IOPaint and LaMa Apache-2.0. The exact big-lama.pt conversion came from the IOPaint maintainer distribution and matched its published checksum. Upstream LaMa links an Apache-2.0 pretrained-weight distribution.
advimman/lama
huggingface.co
Photo credits: Beach photograph by thiago japyassu, Pexels licence, modified for this demonstration. DSCOVR/Falcon 9 photograph by SpaceX, public domain, distributed through scikit-image. No endorsement implied.
pexels.com
pexels.com
flickr.com
What would you clean up?
#GitHub #OpenSource #IOPaint
6 likes · 452 views
Mentions
IOPaint: Erase Objects from Photos, Locally | GitHub Repos You Should Know #015 Paint over an unwanted object and let IOPaint fill the area. Repo: Sanster/IOPaint Classification: Worth Knowing We tested IOPaint 1.6.0 with its LaMa model on the CPU, locally. The examples show actual model outputs from real photos: removing a beach umbrella and shadow, then a tower. Inputs, masks and outputs were retained. Unmasked pixels remained unchanged. No manual output retouching. The mask animation illustrates the actual saved selection. It is an edited demonstration of the local backend, not a recording of the web interface or a processing-speed benchmark. AI estimates replacement pixels; inspect the textures and edges. The photo edits are demonstrations, not documentary evidence of the original scenes. The IOPaint repository is archived. Check compatibility before installing. Other supported models have separate terms. Licences: IOPaint and LaMa Apache-2.0. The exact big-lama.pt conversion came from the IOPaint maintainer distribution and matched its published checksum. Upstream LaMa links an Apache-2.0 pretrained-weight distribution. advimman/lama huggingface.co Photo credits: Beach photograph by thiago japyassu, Pexels licence, modified for this demonstration. DSCOVR/Falcon 9 photograph by SpaceX, public domain, distributed through scikit-image. No endorsement implied. pexels.com pexels.com flickr.com What would you clean up? #GitHub #OpenSource #IOPaint