Skip to content
Other

KrishiLoop AgriTriage

offline model for plant disease diagnosis

2 min readupdated

KrishiLoop AgriTriage is a compact model based on the TinyViT architecture (11 million parameters), built to run directly on smartphones or in the browser. It analyzes plant leaf photos offline and identifies potential diseases across a list of 89 categories. A key feature of the model is its built-in conformal safety gate, which delivers a precise diagnosis, suggests a short-list of options, or sends the photo to an expert depending on confidence levels. The model is useful for mobile app developers and agtech projects that need fast local diagnostics without sending data to a server.

Key features

  • On-device operation: The model weighs just 11.5 MB in ONNX (INT8) format and runs locally.
  • 89 disease classes: Recognizes a wide range of crop diseases.
  • Safety Gate: The safety algorithm reduces the risk of false positives by passing complex cases to an agronomist in time.
  • Interpretability (XAI): Generates a Grad-CAM map to explain which part of the leaf the model focused on.

What you can do

  • Integrate into a mobile app: Use onnxruntime-web or native ONNX Runtime to provide offline assistance to farmers.
  • Create an agtech bot: Apply the model in Telegram bots for preliminary rapid analysis of field photos.
  • Add a verification mechanism: Configure logic to pass uncertain images to human experts.

Example

import json, numpy as np, onnxruntime as ort
from PIL import Image
cfg = json.load(open("config.json"))
s = ort.InferenceSession("model_xai.onnx", providers=["CPUExecutionProvider"])
im = Image.open("leaf.jpg").convert("RGB")
k = cfg["resize"] / min(im.size)                      
im = im.resize((max(224, round(im.width * k)), max(224, round(im.height * k))), Image.Resampling.BILINEAR)
l, t = (im.width - 224) // 2, (im.height - 224) // 2
x = np.asarray(im.crop((l, t, l + 224, t + 224)), np.float32) / 255.0   
out = s.run(None, {s.get_inputs()[0].name: x.transpose(2, 0, 1)[None]})
Read moreDetails
• Top-1 accuracy for the INT8 version is 77.5% (FP32: 77.8%) on the PlantWild test dataset.
• Architecture: TinyViT-11M, distilled from three large teacher models (DINOv2, CLIP, Perception Encoder).
• Out-of-domain performance (on PlantDoc field photos) drops in accuracy without triggering the safety gate.
• Does not process personal data: all processing happens locally on the device.
Discuss Telegram 0 18

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.5 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 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 · BF16 — 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.

    # ABRUBAB/krishiloop-agritriage-tinyvit
    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 ABRUBAB/krishiloop-agritriage-tinyvit --local-dir ~/models/krishiloop-agritriage-tinyvit
    # The model is in ~/models. How to run it is in the model card (link above).
  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 krishiloop-agritriage-tinyvit (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

ABRUBAB/krishiloop-agritriage-tinyvit
Author
ABRUBAB
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

180

Ask AI about this model

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

Comments