File size: 2,504 Bytes
bcd3e7f
 
85cb805
 
 
bcd3e7f
 
 
 
85cb805
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
title: README
emoji: 馃摎
colorFrom: blue
colorTo: gray
sdk: static
pinned: false
---

# MB Mel Dynamics

AI/ML services company in Lithuania. We prepare training data, fine-tune visual and audio models, and deploy them into working systems.

**Website:** [meldynamics.com](https://meldynamics.com) 路 **GitHub:** [MELDYNAMICS](https://github.com/MELDYNAMICS)

## What we publish here

Open datasets and models built on public source material. Each card states what we contributed and credits the original creators.

| Repository | Type | Our contribution | Source material |
|---|---|---|---|
| [polish-metrical-htr-experimental](https://huggingface.co/meldynamics/polish-metrical-htr-experimental) | Model, experimental | Fine-tuned TrOCR line recognizer for 19th-century Polish metrical registers | Base model: Kansallisarkisto/multicentury-htr-model |
| [liepa-tts](https://huggingface.co/datasets/meldynamics/liepa-tts) | Speech dataset | Recovery, alignment, restructuring, and stress enrichment | Vilnius University, LIEPA project |
| [lt-stressed-corpus](https://huggingface.co/datasets/meldynamics/lt-stressed-corpus) | Text dataset | Word-stress enrichment and derived dataset engineering | MATAS v1.0, ALKSNIS v3.0 |
| [liepa-asr](https://huggingface.co/datasets/meldynamics/liepa-asr) | Speech corpus | Hugging Face edition | Vilnius University, LIEPA project |
| [liepa-2](https://huggingface.co/datasets/meldynamics/liepa-2) | Speech corpus | Hugging Face edition | Vilnius University, LIEPA 2 project |
| [liepa-3](https://huggingface.co/datasets/meldynamics/liepa-3) | Speech corpus | Hugging Face edition | Vytautas Magnus University and Vilnius University |

The LIEPA recordings were created by the universities listed above; we did not collect them. The handwriting model is a research model evaluated against automatic teacher transcriptions, not human ground truth.

## Services

- **Visual model fine-tuning:** adapting vision and document models to your images and categories.
- **Audio model fine-tuning:** speech recognition and synthesis for specific languages, speakers, or recording conditions.
- **Dataset collection and management:** collection, cleaning, labelling, alignment, provenance, and versioning.
- **Deployment:** packaging models as services or batch jobs in your environment.

To discuss a project, use the contact form at [meldynamics.com](https://meldynamics.com/#contact).

---

MB Mel Dynamics 路 Company code 306606584 路 Lithuania 路 Established 2023