Skip to content
Other

VML (Visual ML) v2.0.0

Offline Mobile Legends Asset Recognition

2 min readupdated

VML (Visual ML) is a fully offline computer vision model for recognizing visual assets from Mobile Legends: Bang Bang (MLBB). It takes a 224×224 pixel image as input and simultaneously outputs a vector representation (embedding) while classifying the object across two levels of detail, from broad categories down to specific skins or items. Built on the lightweight ConvNeXt-Tiny architecture, the model was distilled (compressed during training without major accuracy loss) from a heavier vit_base_patch16_siglip_224 teacher model by sansaks. Unlike standard classifiers, the release ships with a prebuilt reference vector database for fast similarity search, making it well-suited for analyzing live streams and tournament broadcasts.

Key features

  • Two-level classification: identifies 282 coarse categories (heroes, items, spells, emblems, talents, skills, UI) and 2,794 fine-grained classes, including individual skins across 134 heroes.
  • Built-in database search: ships alongside 4,677 precomputed 768-dimensional normalized embeddings (hero_skin_db.npz) and JSON label-mapping files.
  • ONNX format and offline execution: hero_skin_student.onnx requires no internet connection and runs smoothly via standard ONNX Runtime.
  • Resilient to broadcast artifacts: tested on synthetic tournament frames featuring OBS blur, JPEG compression, overlapping in-game name (IGN) text, and circular icon crops.

What you can do

  • Build a stream overlay or analytics parser to automatically track picks, bans, item builds, and selected skins directly from MLBB live streams and tournament VODs.
  • Add visual similarity search (via 768-dimensional embeddings) to fan databases, wikis, or Telegram bots to identify in-game elements from cropped screenshots.

Example

The model accepts a pixel_values tensor of shape [B, 3, 224, 224] and returns three outputs: embedding (a [B, 768] vector for querying hero_skin_db.npz), coarse_logits ([B, 282] for the base category, such as Skill | ), and fine_logits ([B, 2794] for the exact class, such as | skin_XX).

Read moreDetails
Stress-test accuracy (200 synthetic tournament broadcast frames): Coarse Top-1 — 99.0%, Fine Top-1 — 93.0%, 768-d DB Retrieval Top-1 — 99.5% and Top-5 — 100.0%.
Training dataset: 4,677 standardized PNG assets (134 heroes, icons, portraits, skins, skills, items, spells, emblems, talents, and UI elements).
Included files: hero_skin_student.onnx, hero_skin_student.onnx.data, the hero_skin_db.npz vector database, and JSON label dictionaries.
Discuss Telegram 0

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.

    # sansaks/vml
    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 sansaks/vml --local-dir ~/models/vml
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to vml
  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 vml (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

sansaks/vml
Author
sansaks
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments