wowsuch ai
Live

RL-MINDModel

UHR-BAT2

UHR-BAT2 is a full-weight fine-tune of Qwen3-VL-8B-Instruct for remote-sensing visual question answering. Its released BF16 safetensors checkpoint can be loaded with Transformers without custom remote code.

What it does

Use it to answer questions about remote-sensing images or describe their visible objects and land use.

Availability
Released checkpoint on Hugging Face and ModelScope
Pricing model
Not established
Deployment
Load locally with Transformers; CUDA-enabled PyTorch recommended
Privacy
Not established
Licence
Apache-2.0
Capability
text generationanalysis
Modality
textimagemultimodal
Access
open weights
Deployment
on-premises
Pricing
unknown
Target user
developerresearcher

Evidence

6 claims
  • UHR-BAT2 is fine-tuned for remote-sensing visual question answering. [1]
    **A Qwen3-VL-8B model fine-tuned for remote-sensing visual question answering.**
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The released checkpoint is qwen3vl-8b-geollava-ft3-selected-10k. [1]
    The released checkpoint is `qwen3vl-8b-geollava-ft3-selected-10k`.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The checkpoint can be loaded with Transformers without custom remote code. [1]
    A single image can be processed with the standard Qwen3-VL implementation; custom remote code is not required.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The release includes an inference example demonstrating image-conditioned generation. [1]
    The example demonstrates image-conditioned generation.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The source says counting and small-object localization remain challenging. [1]
    Counting and small-object localization remain challenging.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The source says geographic-disjoint generalization is not established. [1]
    A training-data audit found candidate scene-name/tile matches with RSHR; pixel-level and question-level overlap have not been resolved, so geographic-disjoint generalization is not established.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.

Sources

1
1
Hugging FaceOwner-controlled
https://huggingface.co/RL-MIND/UHR-BAT2 Retrieved 2026-10-28 00:39 UTC
Verified

Versions

2
VersionEffectiveChangeChanged fieldsState
22026-10-15 09:21 UTCmaterial changepricing, deploymentCurrent
12026-10-14 10:20 UTCpublicationNoneCurrent

In briefs

1

Open in the catalog

Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.