Hugging Face's logo

gaborcselle
/
font-identifier

Image Classification
Transformers
TensorBoard
Safetensors
English
resnet
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Metrics Community
font-identifier / README.md
gaborcselle's picture
gaborcselle
Added details to README.md
306ced6
preview code
|
raw
history blame
7.4 kB
metadata
license: mit
base_model: microsoft/resnet-18
tags:
  - generated_from_trainer
datasets:
  - gaborcselle/font-examples
metrics:
  - accuracy
model-index:
  - name: font-identifier
    results:
      - task:
          name: Image Classification
          type: image-classification
        dataset:
          name: imagefolder
          type: imagefolder
          config: default
          split: test
          args: default
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.963265306122449
widget:
  - text: What font is this?
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/ArchitectsDaughter-Regular_1.png
    example_title: Architects Daughter
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Arial%20Bold_39.png
    example_title: Arial Bold
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Courier_28.png
    example_title: Courier
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Helvetica_3.png
    example_title: Helvetica
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/IBMPlexSans-Regular_25.png
    example_title: IBM Plex Sans
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Inter-Regular_43.png
    example_title: Inter
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Lobster-Regular_25.png
    example_title: Lobster
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Merriweather-Regular_1.png
    example_title: Merriweather
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Poppins-Regular_22.png
    example_title: Poppins
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/RobotoMono-Regular_38.png
    example_title: Roboto Mono
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Times_New_Roman_Bold
      Italic_26.png
    example_title: Times New Roman Bold Italic
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Times_New_Roman_Italic_16.png
    example_title: Times New Roman Italic
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/TitilliumWeb-Regular_5.png
    example_title: Titillium Web
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Trebuchet_MS_Italic_47.png
    example_title: Trebuchet MS Italic
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Trebuchet_MS_11.png
    example_title: Trebuchet MS
  - src: >-
      https://huggingface.co/gaborcselle/font-identifier/resolve/main/hf_samples/Verdana_Bold_43.png
    example_title: Verdana Bold
language:
  - en

font-identifier

This model is a fine-tuned version of microsoft/resnet-18 on the imagefolder dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1172
  • Accuracy: 0.9633

Model description

Identify the font used in an image. Visual classifier based on ResNet18.

I built this project in 1 day, with a minute-by-minute journal:

Intended uses & limitations

Identify any of 48 standard fonts from the training data.

Training and evaluation data

Trained and eval'd on the gaborcselle/font-examples dataset (80/20 split).

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 64
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 50

Training results

Training Loss Epoch Step Validation Loss Accuracy
4.0243 0.98 30 3.9884 0.0204
3.7051 1.98 61 3.6012 0.0776
3.2036 2.99 92 2.9556 0.2939
2.6413 4.0 123 2.3054 0.4531
2.1015 4.98 153 1.7366 0.5224
1.6508 5.98 184 1.3509 0.6367
1.3986 6.99 215 1.0938 0.7163
1.1918 8.0 246 0.9012 0.7735
1.0633 8.98 276 0.7464 0.8143
0.8771 9.98 307 0.6569 0.8306
0.8309 10.99 338 0.5536 0.8551
0.7093 12.0 369 0.4795 0.8796
0.6579 12.98 399 0.4176 0.8837
0.5827 13.98 430 0.3888 0.8980
0.5418 14.99 461 0.3255 0.9122
0.5102 16.0 492 0.3139 0.9265
0.472 16.98 522 0.3141 0.9163
0.4273 17.98 553 0.2673 0.9245
0.384 18.99 584 0.2487 0.9265
0.3917 20.0 615 0.2353 0.9388
0.418 20.98 645 0.2113 0.9490
0.3662 21.98 676 0.2095 0.9327
0.3258 22.99 707 0.2139 0.9429
0.3268 24.0 738 0.1962 0.9449
0.3048 24.98 768 0.1935 0.9408
0.2696 25.98 799 0.2112 0.9408
0.2524 26.99 830 0.2310 0.9306
0.2491 28.0 861 0.1827 0.9449
0.2542 28.98 891 0.1720 0.9592
0.2898 29.98 922 0.1605 0.9490
0.2298 30.99 953 0.1326 0.9633
0.2137 32.0 984 0.1438 0.9571
0.2002 32.98 1014 0.1379 0.9551
0.2013 33.98 1045 0.1261 0.9653
0.1862 34.99 1076 0.1674 0.9408
0.1993 36.0 1107 0.1423 0.9571
0.2063 36.98 1137 0.1406 0.9592
0.2088 37.98 1168 0.1717 0.9429
0.1711 38.99 1199 0.1539 0.9510
0.1804 40.0 1230 0.1421 0.9571
0.1793 40.98 1260 0.0765 0.9776
0.2139 41.98 1291 0.1859 0.9449
0.1678 42.99 1322 0.1067 0.9796
0.1675 44.0 1353 0.0985 0.9735
0.1681 44.98 1383 0.1093 0.9653
0.1625 45.98 1414 0.1402 0.9592
0.1987 46.99 1445 0.1250 0.9673
0.1728 48.0 1476 0.1293 0.9633
0.1337 48.78 1500 0.1172 0.9633

Framework versions

  • Transformers 4.36.0.dev0
  • Pytorch 2.0.0
  • Datasets 2.12.0
  • Tokenizers 0.14.1