# Double InfoGAN for Contrastive Analysis Source: [Double InfoGAN for Contrastive Analysis]() Collection index: [llms.txt]() Data updated: 2025-10-26T02:16:25.405Z Authors: Florence Carton, Robin Louiset, Pietro Gori Published: 2024-01-31T12:16:39.000Z ## Abstract Contrastive Analysis (CA) deals with the discovery of what is common and what is distinctive of a target domain compared to a background one. This is of great interest in many applications, such as medical imaging. Current state-of-the-art (SOTA) methods are latent variable models based on VAE (CA-VAEs). However, they all either ignore important constraints or they don't enforce fundamental assumptions. This may lead to sub-optimal solutions where distinctive factors are mistaken for common ones (or viceversa). Furthermore, the generated images have a rather poor quality, typical of VAEs, decreasing their interpretability and usefulness. Here, we propose Double InfoGAN, the first GAN based method for CA that leverages the high-quality synthesis of GAN and the separation power of InfoGAN. Experimental results on four visual datasets, from simple synthetic examples to complex medical images, show that the proposed method outperforms SOTA CA-VAEs in terms of latent separation and image quality. Datasets and code are available online. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Latent Factor Separation on FFHQ No Glasses vs. Glasses](): Latent Factor Separation; FFHQ No Glasses vs. Glasses; C Score: 0.65. - [Common and Salient Separation on BraTS](): Common and Salient Separation; BraTS; Delta (Δ): 0.42. - [Common and Salient Separation on CelebA-HQ Gender](): Common and Salient Separation; CelebA-HQ Gender; Delta Separation Score: 0.4. - [Salient Attribute Swapping on FFHQ Glasses Attribute](): Salient Attribute Swapping; FFHQ Glasses Attribute; Accuracy: 47. - [Latent Factor Separation on FFHQ Head Pose: Frontal vs. Right/Left](): Latent Factor Separation; FFHQ Head Pose: Frontal vs. Right/Left; C Metric: 0.71. - [Attribute Swapping on AFHQ v2 (test)](): Attribute Swapping; AFHQ v2 (test); FID (Y): 126.933. - [Attribute Swapping on BraTS (test)](): Attribute Swapping; BRATS (test); FID-Y: 228.459. - [Attribute Swapping on CelebA-HQ (test)](): Attribute Swapping; CelebA-HQ Gender (test); FID (Y): 88.475. - [Attribute Swapping on CelebA-HQ Smile (test)](): Attribute Swapping; CelebA-HQ Smile (test); FID (Y): 67.475. - [Common and Salient Separation on AFHQ](): Common and Salient Separation; AFHQ; Delta Score: 0.46. - [Common and Salient Separation on CelebA-HQ Smile](): Common and Salient Separation; CelebA-HQ Smile; Separation Delta (Δ): 0.49. - [Reconstruction on AFHQ Background (X) v2 (test)](): Reconstruction; AFHQ Background (X) v2 (test); LPIPS: 0.548. - [Reconstruction on AFHQ Target (Y) v2 (test)](): Reconstruction; AFHQ Target (Y) v2 (test); LPIPS: 0.535. - [Reconstruction on BraTS Background (X) (test)](): Reconstruction; BraTS Background (X) (test); LPIPS: 0.269. - [Reconstruction on BraTS Target (Y) (test)](): Reconstruction; BraTS Target (Y) (test); LPIPS: 0.26. - [Image Reconstruction on CelebA-HQ Gender Background (X) (test)](): Reconstruction; CelebA-HQ Gender Background (X) (test); LPIPS: 0.394. - [Reconstruction on CelebA-HQ Gender Target (Y) (test)](): Reconstruction; CelebA-HQ Gender Target (Y) (test); LPIPS: 0.375. - [Reconstruction on CelebA-HQ Smile Background (test)](): Reconstruction; CelebA-HQ Smile Background (test); LPIPS: 0.336. - [Reconstruction on CelebA-HQ Smile Target (Y) (test)](): Reconstruction; CelebA-HQ Smile Target (Y) (test); LPIPS: 0.308. - [Image Reconstruction on FFHQ (SSIM, LPIPS, ID-Sim)](): Image Reconstruction; FFHQ; SSIM: 32.6. - [Latent Factor Separation on FFHQ (Male vs Female)](): Latent Factor Separation; FFHQ Male vs. Female; C Score: 0.6. - [Latent Factor Separation on FFHQ Smile vs. Non-Smiling](): Latent Factor Separation; FFHQ Smile vs. Non-Smiling; C Score: 0.64. - [Attribute Swapping on FFHQ Swap X to Y high-quality (test)](): Attribute Swapping; FFHQ Swap X to Y high-quality (test); ID Similarity: 0.158. - [Attribute Swapping on FFHQ Swap Y to X high-quality (test)](): Attribute Swapping; FFHQ Swap Y to X high-quality (test); ID-Sim: 0.153. - [Image Reconstruction on FFHQ X dataset high-quality (test)](): Image Reconstruction; FFHQ X dataset high-quality (test); LPIPS: 0.369. - [Image Reconstruction on FFHQ Y dataset high-quality (test)](): Image Reconstruction; FFHQ Y dataset high-quality (test); LPIPS: 0.361. ## Original sources and implementation [Original paper]() [PDF]()