nely-df2d 0.14

2D pose estimation pipeline for tethered Drosophila.

Drosophila 2D Pose

• Load the model.

from model import Drosophila2DPose
from parser import create_parser

checkpoint_path = '/home/user/Desktop/DeepFly3D/weights/sh8_deepfly.tar'
args = create_parser().parse_args('')
model = Drosophila2DPose(checkpoint_path=checkpoint_path, **args.__dict__).cuda()

• Load the data.

from inference import path2inp
from dataset import Drosophila2Dataset
from torch.utils.data import DataLoader

image_path = '/home/user/Desktop/DeepFly3D/data/test/'
inp = path2inp(image_path) # extract list of images under the folder
dat = DataLoader(Drosophila2Dataset(inp), batch_size=8)

• Do the inference.

from inference import inference
points2d = inference(model, dat)