π Awesome OCR multiple programing languages toolkits based on ONNX Runtime, OpenVINO, MNN, PaddlePaddle, TensorRT and PyTorch.
PP-OCRv6_rec_tiny:
model_type: rec
algorithm: SVTR_LCNet
Transform:
Backbone:
name: PPLCNetV4
model_size: tiny
Head:
name: MultiHead
out_channels_list:
CTCLabelDecode: 6906 #'blank' + ...(6904) + ' '
NRTRLabelDecode: 6909
build_gtc_head: false
head_list:
- CTCHead:
Neck:
name: reshape
Head:
mid_channels: 80
use_guide: true
guide_layer_type: ppocrv6_tiny
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 384
max_text_length: 25
PP-OCRv6_rec_small:
model_type: rec
algorithm: SVTR_LCNet
Transform:
Backbone:
name: PPLCNetV4
model_size: small
Head:
name: MultiHead
out_channels_list:
CTCLabelDecode: 18710 #'blank' + ...(18708) + ' '
head_list:
- CTCHead:
Neck:
name: lightsvtr
dims: 120
depth: 2
mlp_ratio: 2.0
local_kernel: 7
Head:
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 384
max_text_length: 25
PP-OCRv6_rec_medium:
model_type: rec
algorithm: SVTR_LCNet
Transform:
Backbone:
name: PPLCNetV4
model_size: medium
Head:
name: MultiHead
out_channels_list:
CTCLabelDecode: 18710 #'blank' + ...(18708) + ' '
head_list:
- CTCHead:
Neck:
name: lightsvtr
dims: 192
depth: 2
mlp_ratio: 4.0
local_kernel: 7
use_guide: false
Head:
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 512
max_text_length: 25
PP-OCRv6_det_tiny:
model_type: det
algorithm: DB
Transform: null
Backbone:
name: PPLCNetV4
det: true
model_size: tiny
Neck:
name: RepLKFPN
out_channels: 64
dilated_kernel_size: 5
shortcut: true
Head:
name: DBHead
k: 50
fix_nan: true
aux_in_channels: 64
PP-OCRv6_det_small:
model_type: det
algorithm: DB
Transform: null
Backbone:
name: PPLCNetV4
det: true
model_size: small
Neck:
name: RepLKFPN
out_channels: 96
dilated_kernel_size: 7
shortcut: true
Head:
name: DBHead
k: 50
fix_nan: true
aux_in_channels: 96
PP-OCRv6_det_medium:
model_type: det
algorithm: DB
Transform: null
Backbone:
name: PPLCNetV4
det: true
model_size: medium
Neck:
name: RepLKPAN
out_channels: 256
intracl: true
Head:
name: DBHead
k: 50
fix_nan: true
aux_in_channels: 256
ch_PP-OCRv5_rec_mobile:
model_type: rec
algorithm: SVTR_LCNet
Transform:
Backbone:
name: PPLCNetV3
scale: 0.95
Head:
name: MultiHead
out_channels_list:
CTCLabelDecode: 18385 #'blank' + ...(18384) + ' '
head_list:
- CTCHead:
Neck:
name: svtr
dims: 120
depth: 2
hidden_dims: 120
kernel_size: [1, 3]
use_guide: True
Head:
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 384
max_text_length: 25
ch_PP-OCRv5_rec_server:
model_type: rec
algorithm: SVTR_HGNet
Transform:
Backbone:
name: PPHGNetV2_B4
text_rec: True
Head:
name: MultiHead
out_channels_list:
CTCLabelDecode: 18385 #'blank' + ...(18384) + ' '
head_list:
- CTCHead:
Neck:
name: svtr
dims: 120
depth: 2
hidden_dims: 120
kernel_size: [1, 3]
use_guide: True
Head:
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 384
max_text_length: 25
ch_PP-OCRv5_det_mobile:
model_type: det
algorithm: DB
Transform: null
Backbone:
name: PPLCNetV3
scale: 0.75
det: True
Neck:
name: RSEFPN
out_channels: 96
shortcut: True
Head:
name: DBHead
k: 50
fix_nan: True
ch_PP-OCRv5_det_server:
model_type: det
algorithm: DB
Transform: null
Backbone:
name: PPHGNetV2_B4
det: True
Neck:
name: LKPAN
out_channels: 256
intracl: true
Head:
name: PFHeadLocal
k: 50
mode: "large"
ch_ppocr_mobile_v2.0_cls_mobile:
model_type: cls
algorithm: CLS
Transform:
Backbone:
name: MobileNetV3
scale: 0.35
model_name: small
Neck:
Head:
name: ClsHead
class_dim: 2
multi_PP-OCRv3_det_mobile:
model_type: det
algorithm: DB
Transform:
Backbone:
name: MobileNetV3
scale: 0.5
model_name: large
disable_se: True
Neck:
name: RSEFPN
out_channels: 96
shortcut: True
Head:
name: DBHead
k: 50
en_PP-OCRv3_det_mobile:
model_type: det
algorithm: DB
Transform:
Backbone:
name: MobileNetV3
scale: 0.5
model_name: large
disable_se: True
Neck:
name: RSEFPN
out_channels: 96
shortcut: True
Head:
name: DBHead
k: 50
en_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR_LCNet
Transform:
Backbone:
name: PPLCNetV3
scale: 0.95
Head:
name: MultiHead
out_channels_list:
CTCLabelDecode: 97 #'blank' + ...(62) + ' '
head_list:
- CTCHead:
Neck:
name: svtr
dims: 120
depth: 2
hidden_dims: 120
kernel_size: [ 1, 3 ]
use_guide: True
Head:
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 384
max_text_length: 25
ch_PP-OCRv4_det_mobile:
model_type: det
algorithm: DB
Transform: null
Backbone:
name: PPLCNetV3
scale: 0.75
det: True
Neck:
name: RSEFPN
out_channels: 96
shortcut: True
Head:
name: DBHead
k: 50
ch_PP-OCRv4_det_server:
model_type: det
algorithm: DB
Transform: null
Backbone:
name: PPHGNet_small
det: True
Neck:
name: LKPAN
out_channels: 256
intracl: true
Head:
name: PFHeadLocal
k: 50
mode: "large"
ch_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR_LCNet
Transform:
Backbone:
name: PPLCNetV3
scale: 0.95
Head:
name: MultiHead
out_channels_list:
CTCLabelDecode: 6625 #'blank' + ...(6623) + ' '
head_list:
- CTCHead:
Neck:
name: svtr
dims: 120
depth: 2
hidden_dims: 120
kernel_size: [ 1, 3 ]
use_guide: True
Head:
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 384
max_text_length: 25
ch_PP-OCRv4_rec_server:
model_type: rec
algorithm: SVTR_HGNet
Transform:
Backbone:
name: PPHGNet_small
Head:
name: MultiHead
out_channels_list:
CTCLabelDecode: 6625 #'blank' + ...(6623) + ' '
head_list:
- CTCHead:
Neck:
name: svtr
dims: 120
depth: 2
hidden_dims: 120
kernel_size: [ 1, 3 ]
use_guide: True
Head:
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 384
max_text_length: 25
chinese_cht_PP-OCRv3_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [1, 2]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 8423
fc_decay: 0.00001
latin_PP-OCRv3_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 187
fc_decay: 0.00001
cyrillic_PP-OCRv3_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 165
fc_decay: 0.00001
arabic_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 164
fc_decay: 0.00001
korean_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 3690
fc_decay: 0.00001
japan_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 4401
fc_decay: 0.00001
ta_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 130
fc_decay: 0.00001
te_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 153
fc_decay: 0.00001
ka_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 155
fc_decay: 0.00001
devanagari_PP-OCRv4_rec_mobile:
model_type: rec
algorithm: SVTR
Transform:
Backbone:
name: MobileNetV1Enhance
scale: 0.5
last_conv_stride: [ 1, 2 ]
last_pool_type: avg
Neck:
name: SequenceEncoder
encoder_type: svtr
dims: 64
depth: 2
hidden_dims: 120
use_guide: True
Head:
name: CTCHead
out_channels: 169
fc_decay: 0.00001