Skip to content

Commit 3d5a7b9

Browse files
zhangyubo0722cuicheng01
zhangyubo0722
and
cuicheng01
authored
add ar models2 (#4604)
Co-authored-by: cuicheng01 <45199522+cuicheng01@users.noreply.github.com>
1 parent e378f36 commit 3d5a7b9

File tree

20 files changed

+1173
-4
lines changed

20 files changed

+1173
-4
lines changed

‎.precommit/check_imports.py‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -96,6 +96,7 @@
9696
"vllm": "vllm",
9797
"xformers": "xformers",
9898
"yarl": "yarl",
99+
"bidi": "python-bidi",
99100
}
100101
MOD_PATTERN = re.compile(
101102
rf"^(?:{'|'.join([re.escape(mod) for mod in MOD_TO_DEP])})(?=\.|$)"

‎docs/module_usage/tutorials/ocr_modules/text_recognition.en.md‎

Lines changed: 45 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -310,6 +310,51 @@ el_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-mo
310310
<td>The Greek recognition model trained based on the PP-OCRv5 recognition model supports recognition of Greek, English, and numbers.</td>
311311
</tr>
312312
<tr>
313+
<td>arabic_PP-OCRv5_mobile_rec</td>
314+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/arabic_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
315+
<td>81.27</td>
316+
<td>-</td>
317+
<td>-</td>
318+
<td>7.6</td>
319+
<td>Ultra-lightweight Arabic character recognition model trained based on the PP-OCRv5 recognition model, supports Arabic letters and number recognition</td>
320+
</tr>
321+
<tr>
322+
<td>cyrillic_PP-OCRv5_mobile_rec</td>
323+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/cyrillic_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
324+
<td>80.27</td>
325+
<td>-</td>
326+
<td>-</td>
327+
<td>7.7</td>
328+
<td>Ultra-lightweight Cyrillic character recognition model trained based on the PP-OCRv5 recognition model, supports Cyrillic letters and number recognition</td>
329+
</tr>
330+
<tr>
331+
<td>devanagari_PP-OCRv5_mobile_rec</td>
332+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/devanagari_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
333+
<td>84.96</td>
334+
<td>-</td>
335+
<td>-</td>
336+
<td>7.5</td>
337+
<td>Ultra-lightweight Devanagari script recognition model trained based on the PP-OCRv5 recognition model, supports Hindi, Sanskrit and other Devanagari letters, as well as number recognition</td>
338+
</tr>
339+
<tr>
340+
<td>te_PP-OCRv5_mobile_rec</td>
341+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/te_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
342+
<td>87.65</td>
343+
<td>-</td>
344+
<td>-</td>
345+
<td>7.5</td>
346+
<td>Ultra-lightweight Telugu script recognition model trained based on the PP-OCRv5 recognition model, supports Telugu script and number recognition</td>
347+
</tr>
348+
<tr>
349+
<td>ta_PP-OCRv5_mobile_rec</td>
350+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/ta_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
351+
<td>94.2</td>
352+
<td>-</td>
353+
<td>-</td>
354+
<td>7.5</td>
355+
<td>Ultra-lightweight Tamil script recognition model trained based on the PP-OCRv5 recognition model, supports Tamil script and number recognition</td>
356+
</tr>
357+
<tr>
313358
<td>korean_PP-OCRv3_mobile_rec</td>
314359
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/korean_PP-OCRv3_mobile_rec_infer.tar">Inference Model</a>/<a href="">Training Model</a></td>
315360
<td>60.21</td>

‎docs/module_usage/tutorials/ocr_modules/text_recognition.md‎

Lines changed: 50 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -319,6 +319,56 @@ el_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model
319319
<td>基于PP-OCRv5识别模型训练得到的希腊语识别模型， 支持希腊语、英文和数字识别</td>
320320
</tr>
321321
<tr>
322+
<td>arabic_PP-OCRv5_mobile_rec</td>
323+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
324+
arabic_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
325+
<td>81.27</td>
326+
<td>-</td>
327+
<td>-</td>
328+
<td>7.6</td>
329+
<td>基于PP-OCRv5识别模型训练得到的超轻量阿拉伯字母识别模型，支持阿拉伯字母、数字识别</td>
330+
</tr>
331+
<tr>
332+
<td>cyrillic_PP-OCRv5_mobile_rec</td>
333+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
334+
cyrillic_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
335+
<td>80.27</td>
336+
<td>-</td>
337+
<td>-</td>
338+
<td>7.7</td>
339+
<td>基于PP-OCRv5识别模型训练得到的超轻量斯拉夫字母识别模型，支持斯拉夫字母、数字识别</td>
340+
</tr>
341+
<tr>
342+
<td>devanagari_PP-OCRv5_mobile_rec</td>
343+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
344+
devanagari_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
345+
<td>84.96</td>
346+
<td>-</td>
347+
<td>-</td>
348+
<td>7.5</td>
349+
<td>基于PP-OCRv5识别模型训练得到的超轻量天城文识别模型，支持印地文、梵文等字母以及数字识别</td>
350+
</tr>
351+
<tr>
352+
<td>te_PP-OCRv5_mobile_rec</td>
353+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
354+
te_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
355+
<td>87.65</td>
356+
<td>-</td>
357+
<td>-</td>
358+
<td>7.5</td>
359+
<td>基于PP-OCRv5识别模型训练得到的超轻量泰卢固文识别模型，支持泰卢固文、数字识别</td>
360+
</tr>
361+
<tr>
362+
<td>ta_PP-OCRv5_mobile_rec</td>
363+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
364+
ta_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
365+
<td>94.2</td>
366+
<td>-</td>
367+
<td>-</td>
368+
<td>7.5</td>
369+
<td>基于PP-OCRv5识别模型训练得到的超轻量泰米尔文识别模型，支持泰米尔文、数字识别</td>
370+
</tr>
371+
<tr>
322372
<td>korean_PP-OCRv3_mobile_rec</td>
323373
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
324374
korean_PP-OCRv3_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/korean_PP-OCRv3_mobile_rec_pretrained.pdparams">训练模型</a></td>

‎docs/pipeline_usage/tutorials/ocr_pipelines/OCR.en.md‎

Lines changed: 45 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -471,6 +471,51 @@ el_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-mo
471471
<td>The Greek recognition model trained based on the PP-OCRv5 recognition model supports recognition of Greek, English, and numbers.</td>
472472
</tr>
473473
<tr>
474+
<td>arabic_PP-OCRv5_mobile_rec</td>
475+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/arabic_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
476+
<td>81.27</td>
477+
<td>-</td>
478+
<td>-</td>
479+
<td>7.6</td>
480+
<td>Ultra-lightweight Arabic character recognition model trained based on the PP-OCRv5 recognition model, supports Arabic letters and number recognition</td>
481+
</tr>
482+
<tr>
483+
<td>cyrillic_PP-OCRv5_mobile_rec</td>
484+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/cyrillic_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
485+
<td>80.27</td>
486+
<td>-</td>
487+
<td>-</td>
488+
<td>7.7</td>
489+
<td>Ultra-lightweight Cyrillic character recognition model trained based on the PP-OCRv5 recognition model, supports Cyrillic letters and number recognition</td>
490+
</tr>
491+
<tr>
492+
<td>devanagari_PP-OCRv5_mobile_rec</td>
493+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/devanagari_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
494+
<td>84.96</td>
495+
<td>-</td>
496+
<td>-</td>
497+
<td>7.5</td>
498+
<td>Ultra-lightweight Devanagari script recognition model trained based on the PP-OCRv5 recognition model, supports Hindi, Sanskrit and other Devanagari letters, as well as number recognition</td>
499+
</tr>
500+
<tr>
501+
<td>te_PP-OCRv5_mobile_rec</td>
502+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/te_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
503+
<td>87.65</td>
504+
<td>-</td>
505+
<td>-</td>
506+
<td>7.5</td>
507+
<td>Ultra-lightweight Telugu script recognition model trained based on the PP-OCRv5 recognition model, supports Telugu script and number recognition</td>
508+
</tr>
509+
<tr>
510+
<td>ta_PP-OCRv5_mobile_rec</td>
511+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/ta_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
512+
<td>94.2</td>
513+
<td>-</td>
514+
<td>-</td>
515+
<td>7.5</td>
516+
<td>Ultra-lightweight Tamil script recognition model trained based on the PP-OCRv5 recognition model, supports Tamil script and number recognition</td>
517+
</tr>
518+
<tr>
474519
<td>korean_PP-OCRv3_mobile_rec</td>
475520
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/korean_PP-OCRv3_mobile_rec_infer.tar">Inference Model</a>/<a href="">Training Model</a></td>
476521
<td>60.21</td>

‎docs/pipeline_usage/tutorials/ocr_pipelines/OCR.md‎

Lines changed: 50 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -448,6 +448,56 @@ el_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model
448448
<td>基于PP-OCRv5识别模型训练得到的希腊语识别模型， 支持希腊语、英文和数字识别</td>
449449
</tr>
450450
<tr>
451+
<td>arabic_PP-OCRv5_mobile_rec</td>
452+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
453+
arabic_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
454+
<td>81.27</td>
455+
<td>-</td>
456+
<td>-</td>
457+
<td>7.6</td>
458+
<td>基于PP-OCRv5识别模型训练得到的超轻量阿拉伯字母识别模型，支持阿拉伯字母、数字识别</td>
459+
</tr>
460+
<tr>
461+
<td>cyrillic_PP-OCRv5_mobile_rec</td>
462+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
463+
cyrillic_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
464+
<td>80.27</td>
465+
<td>-</td>
466+
<td>-</td>
467+
<td>7.7</td>
468+
<td>基于PP-OCRv5识别模型训练得到的超轻量斯拉夫字母识别模型，支持斯拉夫字母、数字识别</td>
469+
</tr>
470+
<tr>
471+
<td>devanagari_PP-OCRv5_mobile_rec</td>
472+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
473+
devanagari_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
474+
<td>84.96</td>
475+
<td>-</td>
476+
<td>-</td>
477+
<td>7.5</td>
478+
<td>基于PP-OCRv5识别模型训练得到的超轻量天城文识别模型，支持印地文、梵文等字母以及数字识别</td>
479+
</tr>
480+
<tr>
481+
<td>te_PP-OCRv5_mobile_rec</td>
482+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
483+
te_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
484+
<td>87.65</td>
485+
<td>-</td>
486+
<td>-</td>
487+
<td>7.5</td>
488+
<td>基于PP-OCRv5识别模型训练得到的超轻量泰卢固文识别模型，支持泰卢固文、数字识别</td>
489+
</tr>
490+
<tr>
491+
<td>ta_PP-OCRv5_mobile_rec</td>
492+
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
493+
ta_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
494+
<td>94.2</td>
495+
<td>-</td>
496+
<td>-</td>
497+
<td>7.5</td>
498+
<td>基于PP-OCRv5识别模型训练得到的超轻量泰米尔文识别模型，支持泰米尔文、数字识别</td>
499+
</tr>
500+
<tr>
451501
<td>korean_PP-OCRv3_mobile_rec</td>
452502
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
453503
korean_PP-OCRv3_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/korean_PP-OCRv3_mobile_rec_pretrained.pdparams">训练模型</a></td>

‎paddlex/configs/modules/text_recognition/arabic_PP-OCRv5_mobile_rec.yaml‎

Lines changed: 39 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,39 @@
1+
Global:
2+
model: arabic_PP-OCRv5_mobile_rec
3+
mode: check_dataset # check_dataset/train/evaluate/predict
4+
dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
5+
device: gpu:0,1,2,3
6+
output: "output"
7+
8+
CheckDataset:
9+
convert:
10+
enable: False
11+
src_dataset_type: null
12+
split:
13+
enable: False
14+
train_percent: null
15+
val_percent: null
16+
17+
Train:
18+
epochs_iters: 20
19+
batch_size: 8
20+
learning_rate: 0.001
21+
pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams
22+
resume_path: null
23+
log_interval: 20
24+
eval_interval: 1
25+
save_interval: 1
26+
27+
Evaluate:
28+
weight_path: "output/best_accuracy/best_accuracy.pdparams"
29+
log_interval: 1
30+
31+
Export:
32+
weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams
33+
34+
Predict:
35+
batch_size: 1
36+
model_dir: "output/best_accuracy/inference"
37+
input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_011_arabic.png"
38+
kernel_option:
39+
run_mode: paddle

‎paddlex/configs/modules/text_recognition/cyrillic_PP-OCRv5_mobile_rec.yaml‎

Lines changed: 39 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,39 @@
1+
Global:
2+
model: cyrillic_PP-OCRv5_mobile_rec
3+
mode: check_dataset # check_dataset/train/evaluate/predict
4+
dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
5+
device: gpu:0,1,2,3
6+
output: "output"
7+
8+
CheckDataset:
9+
convert:
10+
enable: False
11+
src_dataset_type: null
12+
split:
13+
enable: False
14+
train_percent: null
15+
val_percent: null
16+
17+
Train:
18+
epochs_iters: 20
19+
batch_size: 8
20+
learning_rate: 0.001
21+
pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams
22+
resume_path: null
23+
log_interval: 20
24+
eval_interval: 1
25+
save_interval: 1
26+
27+
Evaluate:
28+
weight_path: "output/best_accuracy/best_accuracy.pdparams"
29+
log_interval: 1
30+
31+
Export:
32+
weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams
33+
34+
Predict:
35+
batch_size: 1
36+
model_dir: "output/best_accuracy/inference"
37+
input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_011_cyrillic.png"
38+
kernel_option:
39+
run_mode: paddle

‎paddlex/configs/modules/text_recognition/devanagari_PP-OCRv5_mobile_rec.yaml‎

Lines changed: 39 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,39 @@
1+
Global:
2+
model: devanagari_PP-OCRv5_mobile_rec
3+
mode: check_dataset # check_dataset/train/evaluate/predict
4+
dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
5+
device: gpu:0,1,2,3
6+
output: "output"
7+
8+
CheckDataset:
9+
convert:
10+
enable: False
11+
src_dataset_type: null
12+
split:
13+
enable: False
14+
train_percent: null
15+
val_percent: null
16+
17+
Train:
18+
epochs_iters: 20
19+
batch_size: 8
20+
learning_rate: 0.001
21+
pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams
22+
resume_path: null
23+
log_interval: 20
24+
eval_interval: 1
25+
save_interval: 1
26+
27+
Evaluate:
28+
weight_path: "output/best_accuracy/best_accuracy.pdparams"
29+
log_interval: 1
30+
31+
Export:
32+
weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams
33+
34+
Predict:
35+
batch_size: 1
36+
model_dir: "output/best_accuracy/inference"
37+
input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_012_devanagari.png"
38+
kernel_option:
39+
run_mode: paddle

‎paddlex/configs/modules/text_recognition/ta_PP-OCRv5_mobile_rec.yaml‎

Lines changed: 39 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,39 @@
1+
Global:
2+
model: ta_PP-OCRv5_mobile_rec
3+
mode: check_dataset # check_dataset/train/evaluate/predict
4+
dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
5+
device: gpu:0,1,2,3
6+
output: "output"
7+
8+
CheckDataset:
9+
convert:
10+
enable: False
11+
src_dataset_type: null
12+
split:
13+
enable: False
14+
train_percent: null
15+
val_percent: null
16+
17+
Train:
18+
epochs_iters: 20
19+
batch_size: 8
20+
learning_rate: 0.001
21+
pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams
22+
resume_path: null
23+
log_interval: 20
24+
eval_interval: 1
25+
save_interval: 1
26+
27+
Evaluate:
28+
weight_path: "output/best_accuracy/best_accuracy.pdparams"
29+
log_interval: 1
30+
31+
Export:
32+
weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams
33+
34+
Predict:
35+
batch_size: 1
36+
model_dir: "output/best_accuracy/inference"
37+
input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_008_ta.png"
38+
kernel_option:
39+
run_mode: paddle

0 commit comments

Comments
 (0)