issues
search
huggingface
/
pixparse
Pixel Parsing. A reproduction of OCR-free end-to-end document understanding models with open data
11
stars
1
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Idl pdfa base
#41
danaaubakirova
opened
3 months ago
0
Add handling of IDL&PDFA + generate()
#40
molbap
opened
5 months ago
0
Add OCR eval task, reworked
#39
molbap
closed
8 months ago
0
`image_preprocess` and `anno_preprocess` not used in `hf_dataset` source
#38
molbap
opened
8 months ago
0
Refactor docvqa
#37
molbap
closed
8 months ago
4
Patch size and window size are dropped
#36
molbap
closed
9 months ago
1
Add custom BartDecoder/Attn layer impl with qk norm and F.sdpa support
#35
rwightman
closed
9 months ago
0
add non-cv2 crop_image class and call
#34
molbap
opened
9 months ago
0
add create_transforms and config string to all tasks
#33
molbap
closed
9 months ago
0
Make eval tasks work in refactor
#32
molbap
opened
9 months ago
0
Make finetune tasks run in refactor
#31
molbap
opened
9 months ago
1
add standard _forward methods to tasks
#30
molbap
closed
9 months ago
0
Redundancy between step and _forward
#29
molbap
opened
9 months ago
0
Make all tasks run
#28
molbap
opened
10 months ago
0
Unify +simplify collators API among tasks
#27
molbap
closed
9 months ago
4
unified abstractions for collate functions
#26
molbap
opened
10 months ago
0
Make transforms configurable
#25
molbap
opened
10 months ago
0
A major refactoring to cleanup tech debt, reduce code redundancy
#24
rwightman
opened
10 months ago
0
Pablo/fix tokens docvqa
#23
molbap
opened
10 months ago
0
[Suggestion] Remove crop_margin dependency on cv2
#22
molbap
opened
10 months ago
6
Add dilation / erosion to torchvision based better transforms
#21
rwightman
closed
10 months ago
0
Properly set_interval (epoch) for distributed sampler vs wds
#20
rwightman
closed
10 months ago
0
Adding more transform options
#19
rwightman
closed
10 months ago
0
Pablo/cruller docvqa
#18
molbap
closed
10 months ago
0
Pablo/hotfix
#14
molbap
closed
10 months ago
0
Pablo/task finetune rvlcdip
#13
molbap
closed
11 months ago
2
[Explore] pooling strategies on vision encoder
#12
molbap
opened
11 months ago
0
add try/catch block in case of empty text
#11
molbap
closed
10 months ago
1
[Explore] Curriculum training / progressive resolution
#10
rwightman
opened
12 months ago
2
[BUG] Donut style training (Cruller) w/ vit appears unstable at higher resolution
#9
rwightman
opened
12 months ago
2
[Explore] Where to take vision features?
#8
rwightman
opened
12 months ago
1
[Explore] Vision architecture comparisons at 1280x960
#7
rwightman
opened
12 months ago
1
infer img mean/std from pretrained model
#6
molbap
closed
12 months ago
1
Pablo/add donut comparison task
#5
molbap
closed
11 months ago
6
Missing sanity/unit test for create_loader
#4
molbap
opened
1 year ago
0
Feat/issue 2/find nonempty pages
#3
molbap
closed
1 year ago
3
Support multi-page or valid page selection
#2
molbap
closed
1 year ago
0
Add basic eval_step() on training samples
#1
molbap
closed
1 year ago
6