search

YoungXIAO13 / PoseFromShape

(BMVC 2019) PyTorch implementation of Paper "Pose from Shape: Deep Pose Estimation for Arbitrary 3D Objects"
http://imagine.enpc.fr/~xiaoy/PoseFromShape/
MIT License
170 stars 34 forks source link

no training, code finishes in few mins #3

Closed Not-IITian closed 5 years ago

Not-IITian commented 5 years ago

Namespace(azi_classes=24, batch_size=16, dataset='ObjectNet3D', decrease=200, ele_classes=12, img_feature_dim=1024, inp_classes=24, keypoint=True, lr=0.0001, model=None, mutated=True, n_epoch=300, novel=False, num_render=12, pretrained_resnet=False, print_freq=50, random_range=0, shape='MultiView', shape_feature_dim=256, tour=2, workers=4)

Random Seed: 5015

after running, bash run/train_ObjectNet_3D.sh

Not-IITian commented 5 years ago

also, i see all scripts only use MV shape encoder. Is it possible to run with pointnet encoder?

YoungXIAO13 commented 5 years ago

Namespace(azi_classes=24, batch_size=16, dataset='ObjectNet3D', decrease=200, ele_classes=12, img_feature_dim=1024, inp_classes=24, keypoint=True, lr=0.0001, model=None, mutated=True, n_epoch=300, novel=False, num_render=12, pretrained_resnet=False, print_freq=50, random_range=0, shape='MultiView', shape_feature_dim=256, tour=2, workers=4)

Random Seed: 5015

after running, bash run/train_ObjectNet_3D.sh

Can you check the err_log.txt to see what is the error? And please make sure all the data are properly generated (eg. obj files and multiview rendering images) before training.

For the PointNet part, we'll release the code for training on ObjectNet3D.

Not-IITian commented 5 years ago

ok the issue was related to JPEG loading with library PILLOW. Downgrading it from 6.0 to 5.4.1 reolved the issue..Looking forward to the pointnet encoder code.