Skip to content

update text detection model from DB to PPOCRv3DB #180

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Dec 23, 2023
Merged

update text detection model from DB to PPOCRv3DB #180

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,13 +95,13 @@ Some examples are listed below. You can find more in the directory of each model

![qrcode](./models/qrcode_wechatqrcode/example_outputs/wechat_qrcode_demo.gif)

### Chinese Text detection [DB](./models/text_detection_db/)
### Chinese Text detection [PPOCR-Det](./models/text_detection_ppocr/)

![mask](./models/text_detection_db/example_outputs/mask.jpg)
![mask](./models/text_detection_ppocr/example_outputs/mask.jpg)

### English Text detection [DB](./models/text_detection_db/)
### English Text detection [PPOCR-Det](./models/text_detection_ppocr/)

![gsoc](./models/text_detection_db/example_outputs/gsoc.jpg)
![gsoc](./models/text_detection_ppocr/example_outputs/gsoc.jpg)

### Text Detection with [CRNN](./models/text_recognition_crnn/)

Expand Down
50 changes: 50 additions & 0 deletions benchmark/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,6 +104,10 @@ mean median min input size model
1.19 1.30 1.07 [100, 100] WeChatQRCode with ['detect_2021nov.prototxt', 'detect_2021nov.caffemodel', 'sr_2021nov.prototxt', 'sr_2021nov.caffemodel']
80.97 80.06 73.20 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
80.73 85.47 72.06 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
23.86 24.16 23.26 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
23.94 23.76 23.26 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
26.89 24.78 23.26 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
28.82 29.58 23.26 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
17.97 16.18 12.43 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
19.54 20.66 12.43 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
17.73 24.25 9.65 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -159,6 +163,10 @@ mean median min input size model
5.90 5.90 5.81 [100, 100] WeChatQRCode with ['detect_2021nov.prototxt', 'detect_2021nov.caffemodel', 'sr_2021nov.prototxt', 'sr_2021nov.caffemodel']
2033.55 2454.13 1769.20 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
1896.61 1977.38 1769.20 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
462.50 463.67 456.98 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
462.97 464.33 456.98 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
470.79 464.35 456.98 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
481.71 479.50 456.98 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
237.73 237.57 236.82 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
265.16 270.22 236.82 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
239.69 298.68 198.88 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -215,6 +223,10 @@ mean median min input size model
5.69 5.72 5.66 [100, 100] WeChatQRCode with ['detect_2021nov.prototxt', 'detect_2021nov.caffemodel', 'sr_2021nov.prototxt', 'sr_2021nov.caffemodel']
1070.55 1072.14 1055.67 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
1071.56 1071.38 1055.67 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
238.89 238.22 236.97 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
238.41 240.39 236.97 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
276.96 240.19 236.97 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
304.04 311.21 236.97 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
258.11 258.13 257.64 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
275.27 277.20 257.64 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
254.90 295.88 221.12 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -251,6 +263,10 @@ mean median min input size model
91.40 92.74 89.76 [256, 256] MPPose with ['pose_estimation_mediapipe_2023mar.onnx']
223.24 224.30 216.37 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
223.03 222.28 216.37 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
112.35 111.90 109.99 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
112.68 114.63 109.93 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
183.96 112.72 109.93 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
234.57 249.45 109.93 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
44.24 45.21 41.87 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
45.15 44.15 41.87 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
36.82 46.54 21.75 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -282,6 +298,10 @@ mean median min input size model
91.28 92.89 89.79 [256, 256] MPPose with ['pose_estimation_mediapipe_2023mar.onnx']
254.78 256.13 245.60 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
254.98 255.20 245.60 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
427.53 428.67 425.63 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
427.79 429.28 425.63 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
414.07 429.46 387.26 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
406.10 407.83 383.41 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
33.07 32.88 32.00 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
33.88 33.64 32.00 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
29.32 33.70 20.69 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -332,6 +352,10 @@ mean median min input size model
127.16 173.93 99.77 [256, 256] MPPose with ['pose_estimation_mediapipe_2023mar.onnx']
975.49 977.45 952.43 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
970.16 970.83 928.66 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
238.38 241.90 233.21 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
238.05 236.53 232.05 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
262.58 238.47 232.05 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
280.63 279.26 232.05 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
194.80 195.37 192.65 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
209.49 208.33 192.65 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
192.90 227.02 161.94 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -363,6 +387,8 @@ mean median min input size model
379.46 366.19 360.02 [640, 640] YoloX with ['object_detection_yolox_2022nov_int8.onnx']
33.90 36.32 31.71 [192, 192] MPPalmDet with ['palm_detection_mediapipe_2023feb_int8.onnx']
40.34 41.50 38.47 [128, 256] YoutuReID with ['person_reid_youtu_2021nov_int8.onnx']
162.54 162.78 155.24 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
161.50 160.70 147.69 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
239.68 239.31 236.03 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2022oct_int8.onnx']
199.42 203.20 166.15 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov_int8.onnx']
197.49 169.51 166.15 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2022oct_int8.onnx']
Expand Down Expand Up @@ -413,6 +439,10 @@ mean median min input size model
134.02 136.01 132.06 [256, 256] MPPose with ['pose_estimation_mediapipe_2023mar.onnx']
1441.73 1442.80 1440.26 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
1436.45 1437.89 1430.58 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
360.26 360.82 359.13 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
361.22 361.51 359.13 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
427.85 362.87 359.13 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
475.44 490.06 359.13 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
285.19 284.91 284.69 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
318.96 323.30 284.69 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
289.82 360.87 244.07 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -496,6 +526,10 @@ mean median min input size model
790.98 823.19 755.99 [256, 256] MPPose with ['pose_estimation_mediapipe_2023mar.onnx']
49331.32 49285.30 49210.67 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
49327.34 49489.22 49210.67 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
4422.65 4432.92 4376.19 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
4407.88 4405.92 4353.22 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
3782.89 4404.01 2682.63 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
3472.93 3557.78 2682.63 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
2183.70 2172.36 2156.29 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
2225.19 2222.58 2156.29 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
2214.03 2302.61 2156.29 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -552,6 +586,10 @@ mean median min input size model
41.26 42.74 40.08 [256, 256] MPPose with ['pose_estimation_mediapipe_2023mar.onnx']
384.47 401.25 360.71 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
377.91 381.15 336.30 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
110.51 111.04 107.73 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
110.67 111.54 107.73 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
131.52 111.76 107.73 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
146.42 149.47 107.73 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
68.70 68.63 68.54 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
78.17 80.48 68.54 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
71.42 91.44 61.14 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -608,6 +646,10 @@ mean median min input size model
9.93 9.97 9.82 [100, 100] WeChatQRCode with ['detect_2021nov.prototxt', 'detect_2021nov.caffemodel', 'sr_2021nov.prototxt', 'sr_2021nov.caffemodel']
1914.15 1913.70 1902.25 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
1920.07 1929.80 1902.25 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
495.04 493.75 489.41 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
493.63 491.89 489.41 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
598.94 496.42 489.41 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
667.75 683.91 489.41 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
439.96 441.91 436.49 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
465.56 466.86 436.49 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
431.93 495.94 373.61 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -664,6 +706,10 @@ mean median min input size model
1113.51 1124.83 1106.81 [256, 256] MPPose with ['pose_estimation_mediapipe_2023mar.onnx']
66015.47 65997.60 65993.81 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
66023.14 66034.99 65993.81 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
6094.40 6093.77 6091.85 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
6073.33 6076.77 6055.13 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
5547.32 6057.15 4653.05 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
5284.79 5356.47 4653.05 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
3230.93 3228.61 3228.29 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
3312.02 3323.17 3228.29 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
3262.32 3413.03 3182.11 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down Expand Up @@ -719,6 +765,10 @@ mean median min input size model
548.41 550.86 546.09 [256, 256] MPPose with ['pose_estimation_mediapipe_2023mar.onnx']
34074.19 34077.97 34058.43 [640, 480] DB with ['text_detection_DB_IC15_resnet18_2021sep.onnx']
34073.67 34069.82 34054.29 [640, 480] DB with ['text_detection_DB_TD500_resnet18_2021sep.onnx']
3031.81 3031.79 3030.41 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may.onnx']
3031.41 3031.17 3029.99 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may.onnx']
2638.47 3031.01 1969.10 [640, 480] PPOCRDet with ['text_detection_cn_ppocrv3_2023may_int8.onnx']
2446.99 2500.65 1967.72 [640, 480] PPOCRDet with ['text_detection_en_ppocrv3_2023may_int8.onnx']
1397.09 1396.95 1396.74 [1280, 720] CRNN with ['text_recognition_CRNN_CH_2021sep.onnx']
1428.65 1432.59 1396.74 [1280, 720] CRNN with ['text_recognition_CRNN_CN_2021nov.onnx']
1429.56 1467.34 1396.74 [1280, 720] CRNN with ['text_recognition_CRNN_EN_2021sep.onnx']
Expand Down
Loading