pxthanhlam/shapeclassification

ShapeClassification

Require

• torch >= 1.6
  
• torchvision compatible with torch
  

Usage

• To extarct fetaure

from utils import ShapeRecogModel
model = ShapeRecogModel(model_path = 'path/to/your_model')
img_embd = model.extract_feature(image_path = test_path)
# or
# img_embd = model.extract_feature(image = your_numpy_image)

Model is store at '/data/lamphan/ShapeClassification/checkpoint_0005.pth.tar', ubuntu69

Training

• Code is taken from https://github.com/facebookresearch/moco , with some minor change on augmentation and datal loading part.
• Training command :

python main_moco.py \
  -a resnet50 \
  --lr 0.03 \
  --batch-size 256 \
  --dist-url 'tcp://localhost:10001' --multiprocessing-distributed --world-size 1 --rank 0 --mlp --moco-t 0.2 --aug-plus --cos \
  [your imagenet-folder with train and val folders]

• Please follow the guild in https://github.com/facebookresearch/moco to set your config, compatiple with your resource