Base model: natix-network-org/roadwork
Labels: {0: "None", 1: "Roadwork"}
Training data: subset of natix-network-org/roadwork (validator's own corpus)
Augmentation during training: True (validator's apply_augmentation_by_level)
Train / Val: 4310 / 478
Epochs: 8, LR: 1e-05

Evaluation

Metric	Value
eval_loss	0.3115
eval_accuracy	0.9079
eval_mcc	0.7625
eval_mcc_clamped	0.7625
eval_f1	0.9377
eval_precision	0.9484
eval_recall	0.9272
eval_sim_reward	0.5826
eval_runtime	13.1221
eval_samples_per_second	36.4270
eval_steps_per_second	0.6100
epoch	8.0000

Usage in the SN72 miner

bash

# Option A: pin via env (no config edit needed)
export MINER_LOCAL_MODEL_DIR=/path/to/local/clone

# Option B: point hf_repo at this repo
# Edit base_miner/detectors/configs/ViT_roadwork.yaml:
#   hf_repo: 'vuongnguyen92/roadwork-v6'

Direct inference

python

from transformers import AutoImageProcessor, AutoModelForImageClassification
from PIL import Image
import torch

processor = AutoImageProcessor.from_pretrained("vuongnguyen92/roadwork-v6")
model = AutoModelForImageClassification.from_pretrained("vuongnguyen92/roadwork-v6").eval()

img = Image.open("street.jpg").convert("RGB")
inputs = processor(images=img, return_tensors="pt")
with torch.inference_mode():
    logits = model(**inputs).logits
prob_roadwork = torch.softmax(logits, dim=-1)[0, 1].item()
print(f"P(roadwork) = {prob_roadwork:.3f}")