issues
search
johndpope
/
MegaPortrait-hack
Using Claude Opus to reverse engineer code from MegaPortraits: One-shot Megapixel Neural Head Avatars
https://arxiv.org/abs/2207.07621
39
stars
6
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Feat/celebvhq2 - MultiScalePatchDiscriminator
#47
johndpope
opened
1 day ago
0
replace patchgan -> multiscale patchgan - should improve image quality
#46
johndpope
opened
3 days ago
0
improve tensorboard loss graphs.
#45
johndpope
opened
3 days ago
0
cosface loss bug
#44
hazard-10
closed
3 days ago
3
Feat/sharpen image
#43
johndpope
closed
4 days ago
0
synthesize cross-identity image pairs for the training of our control module
#42
johndpope
opened
1 week ago
0
subsampling on perceptual loss trick from fsrt paper
#41
johndpope
opened
1 week ago
0
Update Dataloader to use voxceleb2 / as well / instead of celebvhq
#40
johndpope
opened
1 week ago
0
Audit Flops - generator / discriminator
#39
johndpope
opened
1 week ago
0
Design for preprocessor at scale / distributed
#38
johndpope
opened
1 week ago
2
Feat/celeb v hq
#37
johndpope
closed
3 days ago
0
Current results of training - epoch 4
#36
johndpope
opened
1 week ago
52
multi-scale patch discriminator + loss
#35
johndpope
closed
1 week ago
2
preprocessing steps - warp / crop / align / mask
#34
johndpope
closed
2 weeks ago
2
DRAFT - Feat/32 cycle consistency
#33
johndpope
closed
1 week ago
3
cycle consistency loss - produce cross-reenacted image: ˆx𝑠∗→𝑑 = Gbase (x𝑠∗ , x𝑑 )
#32
johndpope
closed
2 weeks ago
2
confusoin about renset50 encoder for global descriptor es
#31
hazard-10
closed
1 week ago
5
experiment on reconstructing the input image
#30
flyingshan
closed
2 weeks ago
4
About GPU info
#29
samsara-ku
closed
1 week ago
5
Feat/16 make faster
#28
johndpope
closed
2 weeks ago
1
loss target
#27
ChenyangWang95
closed
1 week ago
7
Reachitect base model to make faster training / use less ram
#26
johndpope
opened
2 weeks ago
7
Implement pre-trained gaze estimation model
#25
johndpope
opened
2 weeks ago
3
Revert "change some model configs"
#24
johndpope
closed
2 weeks ago
0
PR key: zs&es vectors, change model outputs and structure
#23
JackAILab
closed
2 weeks ago
7
change some model configs
#22
JackAILab
closed
2 weeks ago
3
DRAFT - training overhaul - wip
#21
johndpope
closed
2 weeks ago
1
Feat/19 sixdrepnet
#20
johndpope
closed
3 weeks ago
1
HeadPose net - with resnet18 backbone / pitch / yaw / roll
#19
johndpope
closed
2 weeks ago
13
Impossible G3D architecture
#18
robinchm
closed
3 weeks ago
4
Warping twice?
#17
robinchm
closed
3 weeks ago
1
Why not use the newer & improved MetaPortrait instead of reimplementing MegaPortrait?
#16
mrT333
closed
1 week ago
3
WarpingNet / flow fields / grids - reference
#15
johndpope
closed
2 weeks ago
6
Stage 1 - training losses - todo contrastive_loss
#14
johndpope
closed
2 weeks ago
5
fyi - custom resnet50 - reference code fc / maxpool
#13
johndpope
closed
1 week ago
1
Stage 1 - apply_warping_field - vs seems to be wrong size
#12
johndpope
closed
3 weeks ago
2
Stage1 training- warpgenerator blows up OOM
#11
johndpope
closed
3 weeks ago
1
es global descriptor - how to plug this back into Gbase
#10
johndpope
closed
1 month ago
0
One-Shot_Free-View -supplemental
#9
johndpope
closed
1 month ago
0
Genh - high res - stage 2 training - custom resnet50 - residual block. - need to add this
#8
johndpope
opened
1 month ago
0
README
#7
tanshuai0219
closed
1 month ago
2
follow up paper - FSRT: Facial Scene Representation Transformer for Face Reenactment from Factorized Appearance, Head-pose, and Facial Expression Features
#6
johndpope
closed
3 weeks ago
0
warpgenerator - 512-> 568
#5
johndpope
closed
3 weeks ago
17
ResBlock_Custom Mistake?
#4
Kwentar
closed
3 weeks ago
9
Data prep - calculate 3d keypoints
#3
johndpope
closed
1 month ago
2
Training set - LMDBDataset
#2
johndpope
closed
1 week ago
3
Roadmap
#1
johndpope
closed
1 week ago
2