6DRepNet

docker run --gpus all -it --rm \
-v `pwd`:/home/user/workdir \
--shm-size=10g \
ghcr.io/pinto0309/openvino2tensorflow:latest

python train.py

RepVGG Block, identity =  None
RepVGG Block, identity =  None
RepVGG Block, identity =  BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  None
RepVGG Block, identity =  BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  None
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
RepVGG Block, identity =  None
Loading data.
datasets/300W_LP/files.txt
Starting training.
Epoch [1/80], Iter [100/1530] Loss: 0.133982
Epoch [1/80], Iter [200/1530] Loss: 0.081677
Epoch [1/80], Iter [300/1530] Loss: 0.087011
Epoch [1/80], Iter [400/1530] Loss: 0.074134
Epoch [1/80], Iter [500/1530] Loss: 0.080335
Epoch [1/80], Iter [600/1530] Loss: 0.076898
Epoch [1/80], Iter [700/1530] Loss: 0.064817
Epoch [1/80], Iter [800/1530] Loss: 0.066560
Epoch [1/80], Iter [900/1530] Loss: 0.076491
Epoch [1/80], Iter [1000/1530] Loss: 0.054818
Epoch [1/80], Iter [1100/1530] Loss: 0.069879
Epoch [1/80], Iter [1200/1530] Loss: 0.063000
Epoch [1/80], Iter [1300/1530] Loss: 0.051588
Epoch [1/80], Iter [1400/1530] Loss: 0.061051
Epoch [1/80], Iter [1500/1530] Loss: 0.071230
Taking snapshot... None
Epoch [2/80], Iter [100/1530] Loss: 0.046368
Epoch [2/80], Iter [200/1530] Loss: 0.060038
Epoch [2/80], Iter [300/1530] Loss: 0.043622
Epoch [2/80], Iter [400/1530] Loss: 0.054392
...

python convert.py \
output/snapshots/SixDRepNet_1654146629_bs80/_epoch_1.tar \
output.pth

python train.py \
--batch_size 80 \
--num_epochs 100 \
--dataset Pose_300W_LP \
--data_dir datasets/300W_LP_w_masked \
--filename_list datasets/300W_LP_w_masked/files.txt

python train.py \
--batch_size 160 \
--num_epochs 100 \
--dataset Pose_300W_LP \
--data_dir datasets/300W_LP_w_masked \
--filename_list datasets/300W_LP_w_masked/files.txt

python train.py \
--batch_size 768 \
--num_epochs 200 \
--lr 0.0003 \
--dataset Pose_300W_LP \
--data_dir datasets/300W_LP_w_masked \
--filename_list datasets/300W_LP_w_masked/files.txt

EPOCH=37
python train.py \
--batch_size 592 \
--num_epochs 500 \
--lr 0.0003 \
--data_dir datasets/300W_LP_w_masked \
--filename_list datasets/300W_LP_w_masked/files.txt \
--snapshot output/snapshots/SixDRepNet_1660312829_bs56/_epoch_${EPOCH}.pth