YALTAi 2.0.1 hangs forever when segmenting

[johnlockejrr]

YALTAi 2.0.1 hangs forever when segmenting

(train-2.0.1-py3.11) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yaltai kraken -I "*.jpg" --suffix ".xml" segment --yolo runs/detect/train2/weights/best.pt
Loading ANN /home/incognito/YALTAi/train-2.0.1-py3.11/lib/python3.11/site-packages/kraken/blla.mlmodel  Segmenting 234_50a2f_default.jpg

v1.0.2 works:

incognito@DESKTOP-H1BS9PO:~/YALTAi$ source yaltai-1-py3.11/bin/activate
(yaltai-1.0.2-py3.11) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yaltai kraken --device cuda:0 -I "*.jpg" --suffix ".xml" segment --yolo runs/detect/train2/weights/best.pt
scikit-learn version 1.2.2 is not supported. Minimum required version: 0.17. Maximum required version: 1.1.2. Disabling scikit-learn conversion API.
Torch version 2.0.1+cu117 has not been tested with coremltools. You may run into unexpected errors. Torch 2.0.0 is the most recent version that has been tested.
Loading ANN /home/incognito/YALTAi/yaltai-1-py3.11/lib/python3.11/site-packages/kraken/blla.mlmodel     Segmenting
image 1/1 /home/incognito/YALTAi/234_50a2f_default.jpg: 960x736 1 textregion, 24 textlines, 50.6ms
Speed: 6.8ms preprocess, 50.6ms inference, 15.5ms postprocess per image at shape (1, 3, 960, 736)
✓

[johnlockejrr]

Any update?

[PonteIneptique]

I am looking at part of it today.

1. Can you send me a picture or a set of picture you are using ?
2. Have you tried running picture by picture, to see if a specific one is crashing / causing the hanging

[johnlockejrr]

Here is an image: I didn't try to run image by image, trying now and hangs at:

(train-2.0.1-py3.11) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yaltai kraken -I BCUF_Ms._L_2057_330.jpg --suffix ".xml" segment --yolo runs/detect/train2/weights/best.pt
WARNING ⚠️ Ultralytics settings reset to default values. This may be due to a possible problem with your settings or a recent ultralytics package update.
View settings with 'yolo settings' or at '/home/incognito/.config/Ultralytics/settings.yaml'
Update settings with 'yolo settings key=value', i.e. 'yolo settings runs_dir=path/to/dir'.
Loading ANN /home/incognito/YALTAi/train-2.0.1-py3.11/lib/python3.11/site-packages/kraken/blla.mlmodel  Segmenting BCUF_Ms._L_2057_330.jpg

[johnlockejrr]

Since the time I replied here, the segmentation is still hanging...

[PonteIneptique]

can you run yaltai kraken --raise-on-error --verbose ?

[johnlockejrr]

Sure! I'll get back.

[johnlockejrr]

Now dosn't even run.

(train-2.0.1-py3.11) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yaltai kraken --raise-on-error --verbose -I BCUF_Ms._L_2057_330.jpg --suffix ".xml" segment --yolo runs/detect/train2/weights/best.pt

[PonteIneptique]

Can you send me your model somehow ?

[johnlockejrr]

Sure, I upload it now.

best.zip

[PonteIneptique]

I does not hang for me with

yaltai kraken --alto -I ./372407844-ddb4597b-119e-46b5-88da-e6f8346c0be4.jpg -o ".xml" segment -y ./best.pt

[johnlockejrr]

Interesting, which YALTAi version? 2.x or 1.x? Under 1.x works for me, under 2.x hangs

[johnlockejrr]

Under 1.x:

+ convert datasets
+ segment
- train

Under 2.x:

- convert datasets
- segment
+ train

[johnlockejrr]

Just tried on another system:

(yaltai-2.0.2-py3.11) incognito@DESKTOP-NHKR7QL:~/YALTAi$ yaltai kraken --alto -I ./BCUF_Ms._L_2057_330.jpg -o ".xml" segment -y ./best.pt
WARNING ⚠️ Ultralytics settings reset to default values. This may be due to a possible problem with your settings or a recent ultralytics package update.
View settings with 'yolo settings' or at '/home/incognito/.config/Ultralytics/settings.yaml'
Update settings with 'yolo settings key=value', i.e. 'yolo settings runs_dir=path/to/dir'.
Loading ANN /home/incognito/YALTAi/yaltai-2.0.2-py3.11/lib/python3.11/site-packages/kraken/blla.mlmodel Segmenting ./BCUF_Ms._L_2057_330.jpg

Same, hangs.

[PonteIneptique]

I feel like you probably have an issue with your YOLO settings, but I don't understand what it could be, Can you try recreating an environment from scratch just in case ?

[johnlockejrr]

I did create from scratch on both systems, on Python 10 and 11. I do have YOLOv8 installed in the systems under different envs, but I think yolo settings are kept outside the envs in /home/$USER/.config/Ultralytics/settings.yaml can be that?

Mine:

settings_version: 0.0.4
datasets_dir: /home/incognito/datasets
weights_dir: /home/incognito/YALTAi/weights
runs_dir: /home/incognito/YALTAi/runs
uuid: 7a070aad82e209ec117c0be351714f5bda1f17e3e424dd2877f7a44c612126c4
sync: true
api_key: ''
clearml: true
comet: true
dvc: true
hub: true
mlflow: true
neptune: true
raytune: true
tensorboard: true
wandb: true

[PonteIneptique]

I created a new environment and I don't run into the issue. It could be, I guess, the settings.yaml ? Can you do a pip freeze ? Are you using CUDA ?

[johnlockejrr]

Yes, CUDA, under WSL 2 Ubuntu 22.04.5

PyTorch version: 2.1.2+cu121 (torchvision 0.16.2+cu121)
OpenCV version: 4.10.0
OS: Ubuntu 22.04.5 LTS
Python version: 3.11.10
Is CUDA available (PyTorch): Yes
GPU models and configuration: GPU 0: NVIDIA GeForce RTX 4070
Nvidia driver version: 560.94

[PonteIneptique]

OHHH. WSL. Kraken code base is not tested against WSL. It could be that kraken 5.X broke with your install. Can you try to use kraken as a segmenter within the same environment ? I think its 'kraken [samestuff] segment -bl` ?

If that does not work, can you try: pip install yaltai --extra-index-url https://download.pytorch.org/whl/cpu in a new env, to not use CUDA and see what happens ?

[johnlockejrr]

Kraken 4.x and 5.x works flawlessly under WSL 2 I'll try both your sugestions right now.

Segmenting with kraken works in a nanosec.

EDIT:

(yaltai-2.0.2-py3.11) incognito@DESKTOP-NHKR7QL:~/YALTAi$ kraken -I BCUF_Ms._L_2057_330-bw.jpg -o ocr.txt segment
Segmenting      ✓
(yaltai-2.0.2-py3.11) incognito@DESKTOP-NHKR7QL:~/YALTAi$

LAST EDIT:

On CPU works:

(yaltai-2.0.2-py3.11-cpu) incognito@DESKTOP-NHKR7QL:~/YALTAi$ time yaltai kraken --alto -I ./BCUF_Ms._L_2057_330.jpg -o ".xml" segment -y ./best.pt
Loading ANN /home/incognito/YALTAi/yaltai-2.0.2-py3.11-cpu/lib/python3.11/site-packages/kraken/blla.mlmodel     Segmenting ./BCUF_Ms._L_2057_330.jpg
image 1/1 /home/incognito/YALTAi/BCUF_Ms._L_2057_330.jpg: 960x800 1 textregion, 25 textlines, 161.4ms
Speed: 3.7ms preprocess, 161.4ms inference, 0.6ms postprocess per image at shape (1, 3, 960, 800)
✓

real    0m30.502s
user    0m31.600s
sys     0m5.098s

Very strange, trainings does work on GPU and any other envs for other HTR/OCR projects work.

[johnlockejrr]

As far as I remember kraken for segmentation/recognition doesn't use GPU, am I wrong?

[PonteIneptique]

As far as I remember kraken for segmentation/recognition doesn't use GPU, am I wrong? Nope, but from what I get from your output, you are not using kraken segmentation, are you ? You are only using YOLO ? (I see textregion/textlines).

In your previous hanged example, it looks like the issue is on the side of ultralytics/yolo because you do not even get the image 1/1.

1. Can you provide me with a pip freeze please ?
2. Can you run yolo in predict mode alone ?
3. Can you please run the kraken command with --baseline because it looks to me you run into bbox mode

(I am signing off, it's 9 PM here, long past office hours)

[johnlockejrr]

1. pip freeze

(yaltai-2.0.2-py3.10) incognito@DESKTOP-H1BS9PO:~/YALTAi$ pip freeze
aiohappyeyeballs==2.4.3
aiohttp==3.10.8
aiosignal==1.3.1
async-timeout==4.0.3
attrs==24.2.0
certifi==2024.8.30
charset-normalizer==3.3.2
click==8.1.7
contourpy==1.3.0
coremltools==6.3.0
cycler==0.12.1
fast-deskew==1.0
filelock==3.16.1
fonttools==4.54.1
frozenlist==1.4.1
fsspec==2024.9.0
idna==3.10
imageio==2.35.1
importlib_resources==6.4.5
Jinja2==3.1.4
joblib==1.4.2
jsonschema==4.23.0
jsonschema-specifications==2023.12.1
kiwisolver==1.4.7
kraken==5.2.9
lazy_loader==0.4
lightning==2.2.5
lightning-utilities==0.11.7
lxml==5.3.0
markdown-it-py==3.0.0
MarkupSafe==2.1.5
matplotlib==3.9.2
mdurl==0.1.2
mean-average-precision==2021.4.26.0
mpmath==1.3.0
multidict==6.1.0
networkx==3.3
numpy==1.23.5
nvidia-cublas-cu12==12.1.3.1
nvidia-cuda-cupti-cu12==12.1.105
nvidia-cuda-nvrtc-cu12==12.1.105
nvidia-cuda-runtime-cu12==12.1.105
nvidia-cudnn-cu12==8.9.2.26
nvidia-cufft-cu12==11.0.2.54
nvidia-curand-cu12==10.3.2.106
nvidia-cusolver-cu12==11.4.5.107
nvidia-cusparse-cu12==12.1.0.106
nvidia-nccl-cu12==2.18.1
nvidia-nvjitlink-cu12==12.6.68
nvidia-nvtx-cu12==12.1.105
opencv-python==4.10.0.84
packaging==24.1
pandas==2.2.3
pillow==10.4.0
protobuf==3.20.3
psutil==6.0.0
py-cpuinfo==9.0.0
pyarrow==17.0.0
Pygments==2.18.0
pyparsing==3.1.4
python-bidi==0.4.2
python-dateutil==2.9.0.post0
pytorch-lightning==2.4.0
pytz==2024.2
PyWavelets==1.7.0
PyYAML==6.0.2
referencing==0.35.1
regex==2024.9.11
requests==2.32.3
rich==13.9.1
rpds-py==0.20.0
scikit-image==0.21.0
scikit-learn==1.2.2
scipy==1.10.1
seaborn==0.13.2
Shapely==1.8.5.post1
six==1.16.0
sympy==1.13.3
tabulate==0.8.10
thop==0.1.1.post2209072238
threadpoolctl==3.4.0
tifffile==2024.9.20
torch==2.1.2
torchmetrics==1.4.2
torchvision==0.16.2
tqdm==4.66.5
triton==2.1.0
typing_extensions==4.12.2
tzdata==2024.2
ultralytics==8.0.209
urllib3==2.2.3
YALTAi @ file:///home/incognito/YALTAi-2.0.2
yarl==1.13.1

2. how to run yolo in predict mode alone?

(yaltai-2.0.2-py3.10) incognito@DESKTOP-H1BS9PO:~/YALTAi$ kraken --alto -I ./BCUF_Ms._L_2057_330-bw.jpg -o ".xml" segment --baseline
Loading ANN /home/incognito/YALTAi/yaltai-2.0.2-py3.10/lib/python3.10/site-packages/kraken/blla.mlmodel ✓
(yaltai-2.0.2-py3.10) incognito@DESKTOP-H1BS9PO:~/YALTAi$

[PonteIneptique]

It definitely confirms that the issues either lies in the connection between YOLO and Kraken (ie my code) or YOLO itself.

Try to run yolo predict source=BCUF_Ms._L_2057_330-bw.jpg model=best.pt

[johnlockejrr]

(yaltai-2.0.2-py3.10) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yolo predict source=BCUF_Ms._L_2057_330-bw.jpg model=runs/detect/train6/weights/best.pt
Ultralytics YOLOv8.0.209 🚀 Python-3.10.12 torch-2.1.2+cu121 CUDA:0 (NVIDIA GeForce RTX 3060, 12288MiB)
Model summary (fused): 168 layers, 3006038 parameters, 0 gradients, 8.1 GFLOPs

image 1/1 /home/incognito/YALTAi/BCUF_Ms._L_2057_330-bw.jpg: 960x800 (no detections), 74.0ms
Speed: 4.0ms preprocess, 74.0ms inference, 39.2ms postprocess per image at shape (1, 3, 960, 800)
Results saved to runs/detect/predict2
💡 Learn more at https://docs.ultralytics.com/modes/predict

[johnlockejrr]

By the way... I'm not sure if it's a bug or not, when I segment with YOLO only it segments perfectly how I trained, the exact regions... when I segment with YALTAi using the YOLO model, it segment ok but adds more lines without regions names... any idea? I can give you an example if you want.

[PonteIneptique]

OHHHHHHH

image 1/1 /home/incognito/YALTAi/BCUF_Ms._L_2057_330-bw.jpg: 960x800 (no detections), 74.0ms

I am wondering if the issues is that your model does not detect anything in your picture and as a results breaks YALTAi.

Can you zip and send the image, to make sure it's not changed by any form of compression ?

[johnlockejrr]

(yaltai-2.0.2-py3.10) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yolo predict source=3_page-0016.jpg model=runs/detect/train7/weights/best.pt
Ultralytics YOLOv8.0.209 🚀 Python-3.10.12 torch-2.1.2+cu121 CUDA:0 (NVIDIA GeForce RTX 3060, 12288MiB)
Model summary (fused): 168 layers, 3006038 parameters, 0 gradients, 8.1 GFLOPs

image 1/1 /home/incognito/YALTAi/3_page-0016.jpg: 960x704 1 textzone, 9 textlines, 70.4ms
Speed: 4.1ms preprocess, 70.4ms inference, 101.3ms postprocess per image at shape (1, 3, 960, 704)
Results saved to runs/detect/predict4
💡 Learn more at https://docs.ultralytics.com/modes/predict

(yaltai-2.0.2-py3.11-cpu) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yaltai kraken --alto -I ./3_page-0016.jpg -o ".xml" segment -y runs/detect/train7/weights/best.pt
Loading ANN /home/incognito/YALTAi/yaltai-2.0.2-py3.11-cpu/lib/python3.11/site-packages/kraken/blla.mlmodel     Segmenting ./3_page-0016.jpg
image 1/1 /home/incognito/YALTAi/3_page-0016.jpg: 960x704 1 textzone, 9 textlines, 166.4ms
Speed: 3.8ms preprocess, 166.4ms inference, 0.7ms postprocess per image at shape (1, 3, 960, 704)
[10/02/24 12:02:29] WARNING  Polygonizer failed on line 0: all the input arrays must have same number of dimensions, but the array at index 0 has 1 dimension(s) and the array at index 1 has 2 dimension(s)                                                         segmentation.py:783
✓

[PonteIneptique]

Wait so it's not hanging anymore ?

[johnlockejrr]

The YOLO model is ok, detects exactly what I need.

[johnlockejrr]

Wait so it's not hanging anymore ?

On CPU is not hanging.

[johnlockejrr]

The added lines without class names could be from blla.mlmodel? Is that used?

[PonteIneptique]

The added lines without class names could be from blla.mlmodel? Is that used?

The point of YALTAi is to run YOLO Region with BLLA from Kraken, so, yes :)

[johnlockejrr]

Oh... so it adds segmentation from blla also, that's not a problem with PAGE/ALTO, you can select what you need.

[johnlockejrr]

For the issue we closed, is not resolved, at least for me:

(yaltai-2.0.2-py3.10) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yaltai convert alto-to-yolo teyman_alto/*.xml teyman_alto_test --shuffle .1
Using list of inputs.
Found 70 to convert.
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ /home/incognito/YALTAi/yaltai-2.0.2-py3.10/bin/yaltai:8 in <module>                              │
│                                                                                                  │
│   5 from yaltai.cli.yaltai import yaltai_cli                                                     │
│   6 if __name__ == '__main__':                                                                   │
│   7 │   sys.argv[0] = re.sub(r'(-script\.pyw|\.exe)?$', '', sys.argv[0])                         │
│ ❱ 8 │   sys.exit(yaltai_cli())                                                                   │
│   9                                                                                              │
│                                                                                                  │
│ /home/incognito/YALTAi/yaltai-2.0.2-py3.10/lib/python3.10/site-packages/click/core.py:1157 in    │
│ __call__                                                                                         │
│                                                                                                  │
│ /home/incognito/YALTAi/yaltai-2.0.2-py3.10/lib/python3.10/site-packages/click/core.py:1078 in    │
│ main                                                                                             │
│                                                                                                  │
│ /home/incognito/YALTAi/yaltai-2.0.2-py3.10/lib/python3.10/site-packages/click/core.py:1688 in    │
│ invoke                                                                                           │
│                                                                                                  │
│ /home/incognito/YALTAi/yaltai-2.0.2-py3.10/lib/python3.10/site-packages/click/core.py:1688 in    │
│ invoke                                                                                           │
│                                                                                                  │
│ /home/incognito/YALTAi/yaltai-2.0.2-py3.10/lib/python3.10/site-packages/click/core.py:1434 in    │
│ invoke                                                                                           │
│                                                                                                  │
│ /home/incognito/YALTAi/yaltai-2.0.2-py3.10/lib/python3.10/site-packages/click/core.py:783 in     │
│ invoke                                                                                           │
│                                                                                                  │
│ /home/incognito/YALTAi/yaltai-2.0.2-py3.10/lib/python3.10/site-packages/yaltai/cli/yaltai.py:98  │
│ in alto_to_yolo                                                                                  │
│                                                                                                  │
│    95 │   if val:                                                                                │
│    96 │   │   message(f"{len(val)} image for validation.", fg='green')                           │
│    97 │   elif shuffle:                                                                          │
│ ❱  98 │   │   random.shuffle(input_paths)                                                        │
│    99 │   │   val_idx = int(len(input_paths) * shuffle)                                          │
│   100 │   │   message(f"{val_idx+1}/{len(input_paths)} image for validation.", fg='green')       │
│   101                                                                                            │
│                                                                                                  │
│ /usr/lib/python3.10/random.py:394 in shuffle                                                     │
│                                                                                                  │
│   391 │   │   │   for i in reversed(range(1, len(x))):                                           │
│   392 │   │   │   │   # pick an element in x[:i+1] with which to exchange x[i]                   │
│   393 │   │   │   │   j = randbelow(i + 1)                                                       │
│ ❱ 394 │   │   │   │   x[i], x[j] = x[j], x[i]                                                    │
│   395 │   │   else:                                                                              │
│   396 │   │   │   _warn('The *random* parameter to shuffle() has been deprecated\n'              │
│   397 │   │   │   │     'since Python 3.9 and will be removed in a subsequent '                  │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
TypeError: 'tuple' object does not support item assignment

[PonteIneptique]

Could you test with the same image ? AKA BCUF_Ms._L_2057_330-bw.jpg and send me this specific one ?

Because to me

yolo predict source=BCUF_Ms._L_2057_330-bw.jpg model=runs/detect/train6/weights/best.pt

Looks like you had no prediction whatsoever on this specific image

image 1/1 /home/incognito/YALTAi/BCUF_Ms._L_2057_330-bw.jpg: 960x800 (no detections), 74.0ms

And this could be somehow causing issues.

Other point, if you only need YOLO preds, I would use yolo predict and yaltai convert yolo-to-alto

[johnlockejrr]

Ok, wait

[johnlockejrr]

(yaltai-2.0.2-py3.10) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yolo predict source=BCUF_Ms._L_2057_330-bw.jpg model=runs/detect/train3/weights/best.pt
Ultralytics YOLOv8.0.209 🚀 Python-3.10.12 torch-2.1.2+cu121 CUDA:0 (NVIDIA GeForce RTX 3060, 12288MiB)
Model summary (fused): 218 layers, 25840918 parameters, 0 gradients, 78.7 GFLOPs

image 1/1 /home/incognito/YALTAi/BCUF_Ms._L_2057_330-bw.jpg: 960x800 1 textregion, 23 textlines, 81.9ms
Speed: 5.1ms preprocess, 81.9ms inference, 117.0ms postprocess per image at shape (1, 3, 960, 800)
Results saved to runs/detect/predict5
💡 Learn more at https://docs.ultralytics.com/modes/predict

[PonteIneptique]

Is this the same model you sent me ? Because up there, you used a different model (model=runs/detect/train7/weights/best.pt) and (model=runs/detect/train6/weights/best.pt). If you change the params at every test, I am not gonna be able to pin point your issue.

[johnlockejrr]

Can't recall, let me send you this one too, I might have sent you for another language. But nonetheless, it hangs no matter what model.

https://file.io/QHDc71qcCVYP

[PonteIneptique]

I don't have access to a CUDA machine and won't for a few weeks, I can't help you more.

The last thing I would suggest is trying to run

yaltai kraken --verbose --device cuda:0 --raise-on-error ... but that's all I can say

[johnlockejrr]

No problem, I can live with the CPU until then :) You helped much! The only problem is the other issue.

[johnlockejrr]

A side question: with the YOLO model you get 4-point boxes, isn't a way to get multipolygon points like with kraken? I normally train my YOLO models with multi-points like this (label example):

0 0.5693548387096774 0.12229190421892816 0.5693548387096774 0.34207525655644244 0.8157258064516129 0.34207525655644244 0.8157258064516129 0.12229190421892816
1 0.5754032258064516 0.13397947548460662 0.5741935483870968 0.14310148232611175 0.8100806451612903 0.14310148232611175 0.8125 0.13483466362599772 0.8112903225806452 0.12485746864310149 0.5754032258064516 0.12314709236031927 0.5754032258064516 0.13397947548460662
1 0.5754032258064516 0.1507981755986317 0.5754032258064516 0.16077537058152794 0.8125 0.16077537058152794 0.8125 0.1507981755986317 0.8125 0.14139110604332952 0.5754032258064516 0.13996579247434435 0.5754032258064516 0.1507981755986317
1 0.5754032258064516 0.16562143671607754 0.5754032258064516 0.17559863169897377 0.8125 0.17474344355758267 0.8125 0.16562143671607754 0.8125 0.1582098061573546 0.5754032258064516 0.1564994298745724 0.5754032258064516 0.16562143671607754
1 0.5754032258064516 0.18244013683010263 0.5754032258064516 0.19241733181299886 0.8125 0.19241733181299886 0.8125 0.18244013683010263 0.8125 0.17559863169897377 0.5754032258064516 0.17559863169897377 0.5754032258064516 0.18244013683010263
1 0.5741935483870968 0.19811858608893956 0.572983870967742 0.2063854047890536 0.8088709677419355 0.2080957810718358 0.8125 0.1998289623717218 0.8112903225806452 0.18985176738882553 0.5741935483870968 0.18814139110604333 0.5741935483870968 0.19811858608893956
1 0.5754032258064516 0.21493728620296465 0.5741935483870968 0.22405929304446978 0.8100806451612903 0.22405929304446978 0.8125 0.21579247434435575 0.8112903225806452 0.2072405929304447 0.5754032258064516 0.20410490307867732 0.5754032258064516 0.21493728620296465
1 0.5766129032258065 0.2314709236031927 0.5754032258064516 0.23973774230330672 0.8100806451612903 0.23973774230330672 0.8125 0.2323261117445838 0.8112903225806452 0.22234891676168758 0.5766129032258065 0.22063854047890535 0.5766129032258065 0.2314709236031927
1 0.5754032258064516 0.24743443557582667 0.5754032258064516 0.2574116305587229 0.8125 0.2574116305587229 0.8125 0.24743443557582667 0.8125 0.23888255416191562 0.5754032258064516 0.23745724059293044 0.5754032258064516 0.24743443557582667
1 0.5754032258064516 0.2631128848346636 0.5741935483870968 0.27223489167616877 0.8100806451612903 0.27223489167616877 0.8125 0.26396807297605474 0.8112903225806452 0.2557012542759407 0.5754032258064516 0.25057012542759405 0.5754032258064516 0.2631128848346636
1 0.5741935483870968 0.2799315849486887 0.5741935483870968 0.2899087799315849 0.8125 0.2899087799315849 0.8125 0.2799315849486887 0.8125 0.2708095781071836 0.5741935483870968 0.2708095781071836 0.5741935483870968 0.2799315849486887
1 0.5741935483870968 0.29646522234891676 0.572983870967742 0.3055872291904219 0.8088709677419355 0.3055872291904219 0.8112903225806452 0.2973204104903079 0.8100806451612903 0.28734321550741165 0.5741935483870968 0.2864880273660205 0.5741935483870968 0.29646522234891676
1 0.5741935483870968 0.31328392246294184 0.5741935483870968 0.3232611174458381 0.8112903225806452 0.3232611174458381 0.8112903225806452 0.31328392246294184 0.8112903225806452 0.3047320410490308 0.5741935483870968 0.3047320410490308 0.5741935483870968 0.31328392246294184
1 0.572983870967742 0.3289623717217788 0.5717741935483871 0.33808437856328394 0.8088709677419355 0.33808437856328394 0.8112903225806452 0.3298175598631699 0.8100806451612903 0.32069555302166475 0.572983870967742 0.3198403648802737 0.572983870967742 0.3289623717217788

[johnlockejrr]

I don't have access to a CUDA machine and won't for a few weeks, I can't help you more.

The last thing I would suggest is trying to run

yaltai kraken --verbose --device cuda:0 --raise-on-error ... but that's all I can say

(yaltai-2.0.2-py3.10) incognito@DESKTOP-H1BS9PO:~/YALTAi$ yaltai kraken --verbose --device cuda:0 --raise-on-error --alto -I 3_page-0021.jpg -o ".xml" segment -y runs/detect/train7/weights/best.pt

Hangs here doesn't even say it loads the blla.mlmodel

[johnlockejrr]

When training, should I resize the images (and convert the polygons to the new size) to x960 or the trainer takes care of that no matter what size the images have?

[johnlockejrr]

For yaltai convert yolo-to-alto what is the labelmap FILE? How can I get it from segmentation?

[johnlockejrr]

ping