TPIPS — Embedding (late fusion)
Text-conditioned perceptual image similarity, built on
Qwen/Qwen3-VL-Embedding-8B. This repo holds the
embedding checkpoint (one of three TPIPS models, each in its own repo — see
the table at the bottom). Code and full docs: https://github.com/adobe-research/TPIPS.
Late fusion. Each (text, image) is encoded independently into an L2-normalised embedding. The pairwise score is cos(e_a, e_b) (higher = more similar).
Odd-one-out probabilities are a softmax over the three "other-pair" scores
divided by the temperature; 2AFC compares the two reference-candidate scores.
The pairwise score is the model's raw output (temperature is applied at the
probability step).
| Property | Value |
|---|---|
| Base model | Qwen/Qwen3-VL-Embedding-8B |
| Pairwise score | cos(e_a, e_b) |
| Fine-tuning | LoRA (r=16, α=32) on the LLM layers |
| Pooling | last-token |
| Temperature | 0.05 (applied at the probability step) |
Prompt X | Represent the similarity of the image based on X. |
Usage
TPIPS supports Python 3.10 and later. Install matching PyTorch and torchvision builds from the official PyTorch installer, then install TPIPS:
pip install tpips
Start with the recommended embedding model:
import tpips
from PIL import Image
model = tpips.load_model("embedding", device="cuda")
a = Image.open("a.jpg").convert("RGB")
b = Image.open("b.jpg").convert("RGB")
similarity = model.similarity(a, b, factor="lighting") # higher is more similar
distance = model.distance(a, b, factor="lighting") # lower is more similar
The first call downloads the selected TPIPS checkpoint and its Qwen backbone. A CUDA GPU is recommended; FlashAttention is optional.
The TPIPS models
| Model | Repo |
|---|---|
| Embedding (late fusion) | sywang/TPIPS-Embed-Qwen3VL-8B |
| Early Fusion | sywang/TPIPS-EarlyFusion-Qwen3VL-8B |
| Activation Distance | sywang/TPIPS-ActDiff-Qwen3VL-8B |
License
TPIPS is provided under the Adobe Research License for noncommercial research use. See the license for the complete terms.