TLJ Bread Recognition β€” YOLO11s (v2)

CJ AI Campus νŒŒμ΄λ„ ν”„λ‘œμ νŠΈ β€” 뚜레μ₯¬λ₯΄(TLJ) 트레이 μ‚¬μ§„μ—μ„œ λΉ΅ 6쒅을 νƒμ§€ν•˜λŠ” YOLOv11s λͺ¨λΈ.

데이터셋

  • Roboflow tlj-bread-6class v2 (λ²„μŠ€νŠΈμƒ· κ·Έλ£Ή λ‹¨μœ„λ‘œ μž¬λΆ„ν• , train/valid/test κ°„ leakage μ΅œμ†Œν™”)
  • 원본 429μž₯ β†’ train 900 / valid 86 / test 43, 6개 클래슀: choco_swirl_bread, kimchi_croquette, olive_bagel, red_bean_bun, strawberry_donut, twist_donut

ν…ŒμŠ€νŠΈμ…‹ μ„±λŠ₯ (43μž₯, 237개 객체)

μ§€ν‘œ κ°’
mAP@0.5:0.95 95.6%
mAP@0.5 98.6%
Precision 99.0%
Recall 99.3%
μ²˜λ¦¬μ†λ„ (A100, 단일 이미지) ~11.7ms/μž₯

파일

  • best.pt β€” ν•™μŠ΅λœ κ°€μ€‘μΉ˜ (imgsz=640, μ „μ΄ν•™μŠ΅ 100 epoch)
  • test_eval_result_v2.json β€” ν΄λž˜μŠ€λ³„ 상세 μ§€ν‘œ
  • results/ β€” ν•™μŠ΅ 곑선, confusion matrix, PR curve

μ‚¬μš©λ²•

from ultralytics import YOLO
model = YOLO("best.pt")
results = model.predict("tray.jpg")
Downloads last month
13
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support