Matilda - AI Conversion : 2D to 3D

Matilda can produce 3d object by only one image !!

pants

shirts

ring

hat

dress

---

we provice categories

• TOP : shirts
• BTM : pants
• RIN : ring
• DR : dress
• HEA : hat

to be added later

• BRA : bracelet
• NEC : neckless
• BAG : bag
• MAS : mask

Required libraries

• pip install -r requirements.txt
• in requirements.txt, you can look these libraries

Python 3.7
fastapi 
uvicorn[standard] 
ipdb
trimesh
numpy
torch==1.7.1
torchvision==0.8.2
torchaudio==0.7.2
requests
scikit-image
matplotlib
cython==0.29.20
usd-core==22.3
aspose-3d
opencv-python
python-multipart
jwt
PyJWT
httpx

Required Files

├─mask
│  └─DIS
│      └─saved_models
│         {category}.pth
│         - {category} 에는 각 카테고리에 맞는 이름 삽입
│         - ex) 반지 - ring.pth
│         - 각 pth 파일은 google drive에서 얻을 수 있음 
└─predictor
   └─network
      └─models
          isnet.pth
          - google drive 또는 DIS 사이트에서 얻을 수 있음 

Training

• process
• codes and files
  • https://drive.google.com/drive/folders/1b2MMi2FvCjzxXPUgwnFJ0iK3rdLqGHio

Please Clone Extra Loss Function

git clone https://github.com/shubhtuls/PerceptualSimilarity.git

in PerceptualSimilarity.util.util, change code

from skimage.measure import compare_ssim

to

from skimage.metrics import structural_similarity as compare_ssim