Skip to content
Computer vision

camera-island-lora

identifying rare animals and species on camera traps

2 min readupdated

This model tackles two specific tasks: classifying field camera trap images and recognizing endemic Galapagos Islands species that standard tools like SpeciesNet lack labels for. It is built on the lightweight LLaVA-OneVision 0.5B multimodal model (around 0.89B total parameters), fine-tuned with LoRA (a small trainable adapter added on top of frozen weights) on 2,902 labeled frames. It outperforms the base model and standard tools on specialized tasks like telling iguanas apart or spotting invasive rats. The project is geared toward ecologists, wildlife researchers, and developers building conservation systems.

Key features

  • Specialized focus: trained to recognize rare Galapagos species (iguanas, mockingbirds) as well as pests.
  • Lightweight architecture: based on the compact LLaVA-OneVision (0.5B parameters), keeping deployment simple.
  • High accuracy: scores up to 93.3% on tough benchmarks distinguishing similar birds and reptiles.
  • Fast training: the LoRA adapter was trained in just three hours on a single desktop machine (Mac Studio).

What you can do

  • Camera trap analysis: automatically identify animals in field camera footage against a predefined species list.
  • Wildlife conservation: classify endemic Galapagos animals and detect invasive rats inside nature reserves.
  • Ecological monitoring: plug the model into field research pipelines to sort image batches quickly.
  • Custom pipelines: use the repository as a starting point to build your own lightweight multimodal classifiers.

Example

import torch
from PIL import Image
from transformers import AutoProcessor, LlavaOnevisionForConditionalGeneration

repo = "andrewxhill/camera-island-lora"
processor = AutoProcessor.from_pretrained(repo)
model = LlavaOnevisionForConditionalGeneration.from_pretrained(repo, dtype=torch.float32).eval()

question = "Which animal is shown in this crop?\nA. marine iguana\nB. common green iguana"
messages = [{"role": "user", "content": [{"type": "image", "image": Image.open("crop.jpg").convert("RGB")}, {"type": "text", "text": question}]}]
inputs = processor.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt")
out = model.generate(**inputs, do_sample=False, max_new_tokens=64)
print(processor.decode(out[0, inputs["input_ids"].shape[-1]:], skip_special_tokens=True))
Read moreDetails
Model accuracy: 93.33% on birds (mockingbird vs. brown noddy) and 91.69% on marine iguanas.
Training setup: LoRA rank 8 (alpha 16), targeting query and value matrices across all 24 language blocks (540,672 trainable parameters total).
License: CC BY-NC 4.0 (base components use Apache-2.0; iNaturalist dataset carries non-commercial restrictions).
Discuss Telegram 0 3

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 3.3 GB · ready files4.2 GB
  • BF16weights 1.7 GB2.3 GB
  • FP8 / INT8recommendedweights 0.8 GB1.4 GB
  • 4-bitweights 0.5 GB1.0 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # andrewxhill/camera-island-lora
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download andrewxhill/camera-island-lora --local-dir ~/models/camera-island-lora
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to camera-island-lora
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of llava-onevision-qwen2-0.5b-ov (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 12All on Hugging Face
LoRAs and adapters 8All on Hugging Face

About the model

andrewxhill/camera-island-lora
Author
andrewxhill
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

30

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments