Meshy ํ…์Šค์ฒ˜๋ง ๋ถ„๋ฅ˜ ๋ชจ๋ธ

๐Ÿ“Œ ๋ชจ๋ธ ์„ค๋ช…

์ด ๋ชจ๋ธ์€ 3D ๋ชจ๋ธ ์ธ๋„ค์ผ ์ด๋ฏธ์ง€๋ฅผ ์ž…๋ ฅ์œผ๋กœ ๋ฐ›์•„, ์ด๋ฏธ์ง€๊ฐ€ ํ…์Šค์ฒ˜๋ง ๋˜์–ด ์žˆ๋Š”์ง€ ์•„๋‹Œ์ง€๋ฅผ ๋ถ„๋ฅ˜ํ•ฉ๋‹ˆ๋‹ค.

  • ํƒœ์Šคํฌ: ์ด์ง„ ์ด๋ฏธ์ง€ ๋ถ„๋ฅ˜ (Textured vs. Not Textured)
  • ํด๋ž˜์Šค:
    • 0: ํ…์Šค์ฒ˜๋ง ์•ˆ๋จ
    • 1: ํ…์Šค์ฒ˜๋ง ๋จ
  • ๋ชจ๋ธ ๊ตฌ์กฐ: ResNet18 (ImageNet ์‚ฌ์ „ํ•™์Šต ์‚ฌ์šฉ)
  • ์ž…๋ ฅ ์ด๋ฏธ์ง€ ํฌ๊ธฐ: 224 ร— 224 RGB ์ด๋ฏธ์ง€
  • ์ถœ๋ ฅ: 2๊ฐœ์˜ ํด๋ž˜์Šค์— ๋Œ€ํ•œ ๋กœ์ง“๊ฐ’ (์†Œํ”„ํŠธ๋งฅ์Šค ์ „)

๐Ÿ“‚ ํ•™์Šต ๋ฐ์ดํ„ฐ

  • CSV ํŒŒ์ผ: meshy_textured_gold.csv
  • ์ด๋ฏธ์ง€ ์ถœ์ฒ˜: Meshy AI์—์„œ ์ƒ์„ฑ๋œ 3D ์ธ๋„ค์ผ ์ด๋ฏธ์ง€
  • ๋ผ๋ฒจ๋ง ๋ฐฉ๋ฒ•: ์‚ฌ๋žŒ์ด ์ง์ ‘ ํ…์Šค์ฒ˜๋ง ์—ฌ๋ถ€๋ฅผ ๋ณด๊ณ  ์ˆ˜์ž‘์—…์œผ๋กœ ๋ผ๋ฒจ ๋ถ€์—ฌ
  • ์ƒ˜ํ”Œ ์ˆ˜: ์•ฝ 500์žฅ (ํ…์Šค์ฒ˜๋ง / ๋น„ํ…์Šค์ฒ˜๋ง ํด๋ž˜์Šค ๊ท ํ˜• ์žˆ์Œ)

๐Ÿ” ํ‰๊ฐ€ ์ •๋ณด

  • ํ‰๊ฐ€์ง€ํ‘œ: ์ •ํ™•๋„(Accuracy) ๊ธฐ์ค€ ์•ฝ 90% ์ด์ƒ
  • ์†์‹ค ํ•จ์ˆ˜: CrossEntropyLoss
  • ์ตœ์ ํ™” ์•Œ๊ณ ๋ฆฌ์ฆ˜: Adam (lr=1e-4)

๊ฒ€์ฆ ๋ฐ์ดํ„ฐ์…‹์—์„œ๋„ ์•ˆ์ •์ ์ธ ์„ฑ๋Šฅ์„ ๋ณด์˜€์œผ๋ฉฐ, ๋ผ๋ฒจ๊ณผ ์˜ˆ์ธก ๊ฒฐ๊ณผ๋ฅผ ์ด๋ฏธ์ง€๋กœ ์‹œ๊ฐํ™”ํ•˜์—ฌ ์„ฑ๋Šฅ์„ ์ถ”๊ฐ€ ํ™•์ธํ–ˆ์Šต๋‹ˆ๋‹ค.


๐Ÿงช ์‚ฌ์šฉ๋ฒ•

from torchvision import transforms, models
from PIL import Image
import torch

# ๋ชจ๋ธ ๋กœ๋“œ
model = models.resnet18()
model.fc = torch.nn.Linear(model.fc.in_features, 2)
model.load_state_dict(torch.load("pytorch_model.bin", map_location="cpu"))
model.eval()

# ์ „์ฒ˜๋ฆฌ ์ •์˜
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406],
                         [0.229, 0.224, 0.225])
])

# ์˜ˆ์ธก
img = Image.open("example.jpg").convert("RGB")
x = transform(img).unsqueeze(0)
with torch.no_grad():
    logits = model(x)
    pred = logits.argmax(dim=1).item()
    print("ํ…์Šค์ฒ˜๋ง ๋จ" if pred == 1 else "ํ…์Šค์ฒ˜๋ง ์•ˆ๋จ")

๐ŸŽฏ ์‚ฌ์šฉ ๋ชฉ์ 

  • ์ธ๋„ค์ผ ์ด๋ฏธ์ง€๊ฐ€ ํ…์Šค์ฒ˜๋ง๋œ ๊ฒƒ์ธ์ง€ ์ž๋™์œผ๋กœ ๊ตฌ๋ถ„
  • ๋Œ€๊ทœ๋ชจ ๋ฐ์ดํ„ฐ์…‹ ์„ ๋ณ„, ํ’ˆ์งˆ ๊ด€๋ฆฌ, ๋ผ๋ฒจ๋ง ์‹œ๊ฐ„ ๋‹จ์ถ• ๋“ฑ์— ํ™œ์šฉ ๊ฐ€๋Šฅ

๐Ÿท๏ธ ํƒœ๊ทธ

์ด๋ฏธ์ง€ ๋ถ„๋ฅ˜, ์ด์ง„ ๋ถ„๋ฅ˜, 3D ์ธ๋„ค์ผ, ํ…์Šค์ฒ˜๋ง ํŒ๋ณ„, ResNet, Meshy

Downloads last month
2
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support