---
base_model: LiquidAI/LFM2.5-VL-450M
library_name: transformers
tags:
- ship-detection
- satellite-imagery
- vision-language
- lfm2-vl
- ghostwatch
---

# AryanNsc/LMF2.5-VL-Ghost-V1

Fine-tuned [LiquidAI/LFM2.5-VL-450M](https://huggingface.co/LiquidAI/LFM2.5-VL-450M) for maritime
ship detection in satellite imagery. Trained on a combined corpus of HRSC2016,
ShipRSImageNet, and MASATI v2.

The LoRA adapter has been **merged into the base weights** — load it directly:

```python
from transformers import AutoModelForImageTextToText, AutoProcessor

model = AutoModelForImageTextToText.from_pretrained(
    "AryanNsc/LMF2.5-VL-Ghost-V1",
    torch_dtype="bfloat16",
    device_map="auto",
    trust_remote_code=True,
)
processor = AutoProcessor.from_pretrained("AryanNsc/LMF2.5-VL-Ghost-V1", trust_remote_code=True)
```

Part of the [GhostWatch](https://github.com/Guney-olu/ghostwatch) dark-vessel detection system.