reupdate

config/base.yml

+# change to list chars of your dataset or use default vietnamese chars
+vocab: 'aAàÀảẢãÃáÁạẠăĂằẰẳẲẵẴắẮặẶâÂầẦẩẨẫẪấẤậẬbBcCdDđĐeEèÈẻẺẽẼéÉẹẸêÊềỀểỂễỄếẾệỆfFgGhHiIìÌỉỈĩĨíÍịỊjJkKlLmMnNoOòÒỏỎõÕóÓọỌôÔồỒổỔỗỖốỐộỘơƠờỜởỞỡỠớỚợỢpPqQrRsStTuUùÙủỦũŨúÚụỤưƯừỪửỬữỮứỨựỰvVwWxXyYỳỲỷỶỹỸýÝỵỴzZ0123456789!"#$%&''()*+,-./:;<=>?@[\]^_`{|}~ '
+
+# cpu, cuda, cuda:0
+device: cuda:0
+
+seq_modeling: transformer
+transformer:  
+    d_model: 256
+    nhead: 8
+    num_encoder_layers: 6
+    num_decoder_layers: 6
+    dim_feedforward: 2048
+    max_seq_length: 1024
+    pos_dropout: 0.1
+    trans_dropout: 0.1
+
+optimizer:
+    max_lr: 0.0003 
+    pct_start: 0.1
+
+trainer:
+    batch_size: 32
+    print_every: 200
+    valid_every: 4000
+    iters: 100000
+    # where to save our model for prediction
+    export: ./weights/transformerocr.pth
+    checkpoint: ./checkpoint/transformerocr_checkpoint.pth
+    log: ./train.log
+    # null to disable compuate accuracy, or change to number of sample to enable validiation while training
+    metrics: null
+
+dataset:    
+    # name of your dataset
+    name: data
+    # path to annotation and image
+    data_root: ./img/
+    train_annotation: annotation_train.txt
+    valid_annotation: annotation_val_small.txt
+    # resize image to 32 height, larger height will increase accuracy
+    image_height: 32
+    image_min_width: 32
+    image_max_width: 512
+
+dataloader:
+    num_workers: 3
+    pin_memory: True
+
+aug:
+    image_aug: true
+    masked_language_model: true
+
+predictor:
+    # disable or enable beamsearch while prediction, use beamsearch will be slower
+    beamsearch: False
+
+quiet: False 
\ No newline at end of file

config/vgg-seq2seq.yml

+pretrain: 
+    id_or_url: 1nTKlEog9YFK74kPyX0qLwCWi60_YHHk4
+    md5: efcabaa6d3adfca8e52bda2fd7d2ee04
+    cached: /tmp/tranformerorc.pth
+
+# url or local path
+weights: https://drive.google.com/uc?id=1nTKlEog9YFK74kPyX0qLwCWi60_YHHk4
+
+backbone: vgg19_bn
+cnn:
+    # pooling stride size
+    ss:
+        - [2, 2]
+        - [2, 2]
+        - [2, 1]
+        - [2, 1]
+        - [1, 1]         
+    # pooling kernel size 
+    ks:
+        - [2, 2]
+        - [2, 2]
+        - [2, 1]
+        - [2, 1]
+        - [1, 1]
+    # dim of ouput feature map
+    hidden: 256
+
+seq_modeling: seq2seq
+transformer:
+    encoder_hidden: 256
+    decoder_hidden: 256
+    img_channel: 256
+    decoder_embedded: 256
+    dropout: 0.1
+
+optimizer:
+    max_lr: 0.001
+    pct_start: 0.1
\ No newline at end of file

config/vgg-transformer.yml

+# for training
+pretrain: https://vocr.vn/data/vietocr/vgg_transformer.pth
+
+# url or local path (predict)
+weights: https://vocr.vn/data/vietocr/vgg_transformer.pth
+
+backbone: vgg19_bn
+cnn:
+    pretrained: True
+    # pooling stride size
+    ss:
+        - [2, 2]
+        - [2, 2]
+        - [2, 1]
+        - [2, 1]
+        - [1, 1]         
+    # pooling kernel size 
+    ks:
+        - [2, 2]
+        - [2, 2]
+        - [2, 1]
+        - [2, 1]
+        - [1, 1]
+    # dim of ouput feature map
+    hidden: 256
+
+seq_modeling: transformer
+transformer:
+    encoder_hidden: 256
+    decoder_hidden: 256
+    img_channel: 256
+    decoder_embedded: 256
+    dropout: 0.1
+
+optimizer:
+    max_lr: 0.001
+    pct_start: 0.1
\ No newline at end of file

inferonnx_croptext.ipynb

@@ -214,7 +214,7 @@
     "    im = cv2.copyMakeBorder(im, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color)  # add border\n",
     "    return im, r, (dw, dh)\n",
     "\n",
-    "names = ['id', 'name', 'birth', 'sex', 'nation', 'origin', 'origin1', 'r1', 'r2']\n",
+    "names = [\"birth\", \"id\", \"name\", \"nation\", \"origin\", \"origin1\", \"r1\", \"r2\", \"sex\"]\n",
     "colors = {name:[random.randint(0, 255) for _ in range(3)] for i,name in enumerate(names)}\n",
     "\n",
     "img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",

infertorchscript_ocr.ipynb

@@ -366,9 +366,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "model_cnn = torch.jit.load(\"/home/anhcd/Projects/dac_vietocr/ConvertVietOcr2Onnx/weight/cnn.cpu.torchscript\").to(device)\n",
-    "model_encoder = torch.jit.load(\"/home/anhcd/Projects/dac_vietocr/ConvertVietOcr2Onnx/weight/encoder.cpu.torchscript\").to(device)\n",
-    "model_decoder = torch.jit.load(\"/home/anhcd/Projects/dac_vietocr/ConvertVietOcr2Onnx/weight/decoder.cpu.torchscript\").to(device)"
+    "model_cnn = torch.jit.load(\"/home/anhcd/Projects/dac_vietocr/ConvertVietOcr2Onnx/weight/cnn.torchscript\").to(device)\n",
+    "model_encoder = torch.jit.load(\"/home/anhcd/Projects/dac_vietocr/ConvertVietOcr2Onnx/weight/encoder.torchscript\").to(device)\n",
+    "model_decoder = torch.jit.load(\"/home/anhcd/Projects/dac_vietocr/ConvertVietOcr2Onnx/weight/decoder.torchscript\").to(device)"
    ]
   },
   {