font-identifier

This model is a fine-tuned version of microsoft/resnet-18 on the imagefolder dataset. Result: Loss: 0.1172; Accuracy: 0.9633

Try with any screenshot of a font, or any of the examples in the 'samples' subfolder of this repo.

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 on Twitter/X, on Pebble.social, and on Threads.net.

The code used to build this model is in this github rep

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

Confusion Matrix

Confusion matrix on test data.

Framework versions

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