From 4c5406c8334a4db287446ce54ce509a4bec6b2b9 Mon Sep 17 00:00:00 2001 From: Shobhita Sundaram Date: Tue, 9 Jan 2024 15:22:47 -0500 Subject: [PATCH 1/3] List of upcoming updates --- README.md | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/README.md b/README.md index ef8fe50..9bade93 100644 --- a/README.md +++ b/README.md @@ -16,6 +16,12 @@ Current metrics for perceptual image similarity operate at the level of pixels a DreamSim is a new metric for perceptual image similarity that bridges the gap between "low-level" metrics (e.g. LPIPS, PSNR, SSIM) and "high-level" measures (e.g. CLIP). Our model was trained by concatenating CLIP, OpenCLIP, and DINO embeddings, and then finetuning on human perceptual judgements. We gathered these judgements on a dataset of ~20k image triplets, generated by diffusion models. Our model achieves better alignment with human similarity judgements than existing metrics, and can be used for downstream applications such as image retrieval. +## 🕰️ Coming soon: +* JND Dataset release +* Distilled DreamSim models (i.e. smaller models distilled from the main ensemble) +* DreamSim variants trained for higher resolutions +* Compatibility with the most recent version of PeFT + ## 🚀 Updates **7/14/23:** Released three variants of DreamSim that each only use one finetuned ViT model instead of the full ensemble. These single-branch models provide a ~3x speedup over the full ensemble. From 008b890fa8377944003bf4ba6c62376a27805443 Mon Sep 17 00:00:00 2001 From: Shobhita Sundaram Date: Tue, 9 Jan 2024 15:23:01 -0500 Subject: [PATCH 2/3] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 9bade93..30f5a16 100644 --- a/README.md +++ b/README.md @@ -16,7 +16,7 @@ Current metrics for perceptual image similarity operate at the level of pixels a DreamSim is a new metric for perceptual image similarity that bridges the gap between "low-level" metrics (e.g. LPIPS, PSNR, SSIM) and "high-level" measures (e.g. CLIP). Our model was trained by concatenating CLIP, OpenCLIP, and DINO embeddings, and then finetuning on human perceptual judgements. We gathered these judgements on a dataset of ~20k image triplets, generated by diffusion models. Our model achieves better alignment with human similarity judgements than existing metrics, and can be used for downstream applications such as image retrieval. -## 🕰️ Coming soon: +## 🕰️ Coming soon * JND Dataset release * Distilled DreamSim models (i.e. smaller models distilled from the main ensemble) * DreamSim variants trained for higher resolutions From 99222ad4cd4512e975721665336fa8c795990ec3 Mon Sep 17 00:00:00 2001 From: Shobhita Sundaram Date: Tue, 9 Jan 2024 15:37:22 -0500 Subject: [PATCH 3/3] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 30f5a16..9944e47 100644 --- a/README.md +++ b/README.md @@ -20,7 +20,7 @@ DreamSim is a new metric for perceptual image similarity that bridges the gap be * JND Dataset release * Distilled DreamSim models (i.e. smaller models distilled from the main ensemble) * DreamSim variants trained for higher resolutions -* Compatibility with the most recent version of PeFT +* Compatibility with the most recent version of PEFT ## 🚀 Updates