File size: 5,583 Bytes
05e2184
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ff7fe7e
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
---
license: mit
tags:
- multimodal
- embeddings
datasets:
- ituperceptron/image-captioning-turkish
- dogukanvzr/ml-paraphrase-tr
library_name: pytorch
language:
- tr
base_model:
- newmindai/modernbert-base-tr-uncased-allnli-stsb
- facebook/dinov2-base
---

# Turkish Multimodal Embedding Model

This repository contains a **contrastively trained Turkish multimodal embedding model**, combining a text encoder and a vision encoder with projection heads.  
The model is trained entirely on **Turkish datasets** (image–caption and paraphrase), making it specifically tailored for Turkish multimodal applications.

## Model Summary
- **Text encoder**: `newmindai/modernbert-base-tr-uncased-allnli-stsb`
- **Vision encoder**: `facebook/dinov2-base`
- **Dimensions**: `text_dim=768`, `image_dim=768`, `embed_dim=768`
- **Projection dropout**: fixed at `0.4` (inside `ProjectionHead`)
- **Pooling**: mean pooling over tokens (`use_mean_pooling_for_text=True`)
- **Normalize outputs**: `{normalize}`
- **Encoders frozen during training?**: `{frozen}` (this release was trained with encoders **NOT frozen**)
- **Language focus**: Turkish (both text and image–caption pairs are fully in Turkish)

## Training Strategy (inspired by JINA-CLIP-v2 style)
- The model was trained jointly with **image–text** and **text–text** pairs using a **bidirectional contrastive loss** (InfoNCE/CLIP-style).
- For **image–text**, standard CLIP-style training with **in-batch negatives** was applied.
- For **text–text**, only **positive paraphrase pairs (label=1)** were used, with in-batch negatives coming from other samples.
- This follows the general training philosophy often seen in Jina’s multimodal work, but in a **simplified single-stage setup** (without the 3-stage curriculum).

## Datasets
- **Image–Text**: [`ituperceptron/image-captioning-turkish`](https://huggingface.co/datasets/ituperceptron/image-captioning-turkish)  
- **Text–Text (Paraphrase)**: [`dogukanvzr/ml-paraphrase-tr`](https://huggingface.co/datasets/dogukanvzr/ml-paraphrase-tr)  

> Both datasets are in Turkish, aligning the model’s embedding space around Turkish multimodal signals.  
> Please check each dataset’s license and terms before downstream use.

## Files
- `pytorch_model.bin` — PyTorch `state_dict`
- `config.json` — metadata (encoder IDs, dimensions, flags)
- `model.py` — custom model classes (required to load)
- (This README is the model card.)

## Evaluation Results
**Dataset:** Test split created from [`ituperceptron/image-captioning-turkish`](https://huggingface.co/datasets/ituperceptron/image-captioning-turkish)

### Image-Text
**Average cosine similarity:** 0.7934

**Recall@K**
<table>
<tr><th>Direction</th><th>R@1</th><th>R@5</th><th>R@10</th></tr>
<tr><td>Text → Image</td><td>0.9365</td><td>0.9913</td><td>0.9971</td></tr>
<tr><td>Image → Text</td><td>0.9356</td><td>0.9927</td><td>0.9958</td></tr>
</table>

<details>
<summary>Raw metrics (JSON)</summary>

```json
{
    "avg_cosine_sim": 0.7934404611587524,
    "recall_text_to_image": {
        "R@1": 0.936458564763386,
        "R@5": 0.9913352588313709,
        "R@10": 0.9971117529437903
    },
    "recall_image_to_text": {
        "R@1": 0.9355698733614752,
        "R@5": 0.9926682959342369,
        "R@10": 0.9957787158409243
    }
}
```
</details>

### Text-Text
**Average cosine similarity:** 0.7599

**Recall@K**
<table>
<tr><th>Direction</th><th>R@1</th><th>R@5</th><th>R@10</th></tr>
<tr><td>Text → Text</td><td>0.7198</td><td>0.9453</td><td>0.9824</td></tr>
</table>

<details>
<summary>Raw metrics (JSON)</summary>
  
```json
{
    "avg_cosine_sim": 0.7599335312843323,
    "recall_text_to_text": {
        "R@1": 0.719875500222321,
        "R@5": 0.9453090262338817,
        "R@10": 0.9824366385060027
    }
}
```
</details>

## Loading & Usage
```python
import torch
import torch.nn.functional as F
from transformers import AutoModel, AutoTokenizer, AutoImageProcessor
from PIL import Image

device = "cuda" if torch.cuda.is_available() else "cpu"

model_name = "utkubascakir/MultiEmbedTR"

model = AutoModel.from_pretrained(model_name, trust_remote_code=True).to(device)
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
image_processor = AutoImageProcessor.from_pretrained(model_name, trust_remote_code=True)

model.eval()

# Text Embedding
texts = ["yeşil arka planlı bir kedi", "kumsalda bir köpek"]
text_inputs = tokenizer(
    texts,
    padding=True,
    truncation=True,
    return_tensors="pt"
).to(device)

with torch.no_grad():
    text_embeds = model.encode_text(
        input_ids=text_inputs["input_ids"],
        attention_mask=text_inputs["attention_mask"]
    )  

print("Text embeddings shape:", text_embeds.shape)

# Image Embedding
img = Image.open("kedi.jpg").convert("RGB")
image_inputs = image_processor(
    images=img,
    return_tensors="pt"
).to(device)

with torch.no_grad():
    image_embeds = model.encode_image(
        pixel_values=image_inputs["pixel_values"]
    ) 

print("Image embeddings shape:", image_embeds.shape)

similarity = F.cosine_similarity(text_embeds, image_embeds)
print("Cosine similarity:", similarity)
```
  
## Limitations & Intended Use
This release provides a **Turkish multimodal embedding model**, trained to produce aligned vector representations for text and images.  
It has not been tested for specific downstream tasks (e.g., retrieval, classification).  
No guarantees for bias/toxicity; please evaluate on your own target domain.

## Citation
If you use this model, please cite this repository.