AI Paper Atlas

Research map / Image generation / Image restoration

FiLM-CAM: Keyed Feature Modulation for Conditional-Access Watermarking

Yash Kulthe, Vishal Asnani, Shruti Agarwal, Andrew Gilbert, John Collomosse

NeurIPS 2026 · Image generation › Image restoration · approach: Image harmonization and color transformation · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes FiLM-CAM, a conditional-access image watermarking method in which a secret key drives feature-wise linear modulation of the watermark signal's internal representation, so that encoding, decoding, and removal all require the matching key.

Problem

Image watermarking typically lets anyone with the detector read or remove a watermark; the paper targets conditional access, where encoding, decoding and removal should each be restricted to authorized key holders.

Approach

A secret key conditions a FiLM mechanism that modulates the internal feature representation of the watermark before embedding, coupling the payload to the key at the feature level rather than the pixel level.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Replace FiLM conditioning with direct key input to the embedder (ablation)FiLM-CAM · DIV2Kbit accuracy: 0.85 -> 0.99Improved
Add FiLM conditioning (ablation)FiLM-CAM · DIV2Kremoval PSNR (dB): 47.3 -> 55.4Improved
Add the image-dependent gain term (ablation)FiLM-CAM · DIV2Kremoval PSNR (dB): 44.0 -> 55.4Improved
Add the gain term (ablation)FiLM-CAM · DIV2Kbit accuracy: 0.99 -> 0.99No clear effect
Use logit-space composition instead of additive composition (ablation)FiLM-CAM · DIV2Kremoval PSNR (dB): 54.1 -> 55.4Improved
Use logit-space composition instead of additive composition (ablation)FiLM-CAM · COCOremoval PSNR (dB): 55.5 -> 59.3Improved
Use logit-space composition instead of additive composition (ablation)FiLM-CAM · DIV2Kwatermarked PSNR (dB): 47.0 -> 46.7No clear effect
Use ConvNeXt-Base extractor instead of ResNet-50 (ablation)FiLM-CAM encoder · DIV2Kwatermarked PSNR (dB): 37.5 -> 46.7Improved
Add noise layer during training (ablation)FiLM-CAM · COCOwatermarked PSNR (dB): 48.2 -> 40.2Hurt
Add noise layer during training (ablation)FiLM-CAM · Wrong-key removalpost-removal bit accuracy: 0.81 wrong-key survivalImproved
Increase iterative refinement steps beyond 20 for removal (ablation)FiLM-CAM · removal PSNRremoval PSNRNo clear effect
Repeatedly re-embed watermarks without removal (ablation)FiLM-CAM · Repeated Watermark EmbeddingPSNR and bit accuracyHurt

Experimental setup

Builds on: FiLM

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy