Skip to content

ObjectDrop

Remove or insert objects with realistic shadows and reflections

Visit official site

What objectdrop is

The method is a supervised approach for photorealistic object removal and insertion. It draws on a counterfactual dataset and self supervision to deal with occlusions, shadows and reflections.

What you can do with it

  • Remove a car from a landscape photo along with its scene effects
  • Insert a desk into an office photo with plausible shadows and occlusion
  • Move a sofa in a room and repair the floor texture at the old spot

What to watch out for

  • Removal relies on fine tuning a diffusion model on a small captured dataset
  • Insertion first synthesizes a counterfactual dataset from the removal model

Pros & cons

✓ What we like

  • Scene effects such as shadows are handled together with the object
  • Insertion is trained on a synthesized large scale counterfactual dataset

! What to watch out for

  • Removal relies on fine tuning a diffusion model on a small captured dataset
  • Insertion first synthesizes a counterfactual dataset from the removal model

FAQ

What operations are listed?

The material lists object removal, object insertion and object movement.

What scene effects are considered?

Occlusion, shadows and reflections are named as effects to handle.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review