Add fine-tuned Comic Sans detector

.ipynb_checkpoints/README-checkpoint.md

@@ -0,0 +1,69 @@
+# Comic Sans Detector
+
+This repository contains a fine-tuned ResNet-18 model, specifically trained to detect whether an image contains Comic Sans font. It is a fine-tuning of a previously fine-tuned font classification model, based on the ResNet-18 foundation model.
+
+## Repository Contents
+
+- **`comic-detector.ipynb`**: A notebook that demonstrates the training and evaluation process for the Comic Sans detector using the fine-tuned ResNet-18 model.
+- **`image-format-generalizer.ipynb`**: A utility notebook for preparing and normalizing image datasets, ensuring consistent formatting across `/data` folders.
+
+## Dataset Structure (Not Included)
+
+The dataset used for training and evaluation should follow this structure:
+```
+/data
+├── comic/
+│   ├── image1.jpg
+│   ├── image2.png
+│   └── ...
+├── not-comic/
+│   ├── image1.jpg
+│   ├── image2.png
+│   └── ...
+```
+- **`comic/`**: Contains images labeled as featuring Comic Sans font.
+- **`not-comic/`**: Contains images labeled as not featuring Comic Sans font.
+
+⚠️ The dataset itself is not included in this repository. You must prepare and structure your dataset as described.
+
+## How to Use
+
+### 1. Clone the Repository
+```bash
+git clone https://huggingface.co/your-username/comic-sans-detector
+cd comic-sans-detector
+```
+
+### 2. Prepare the Dataset
+Ensure your dataset is properly structured under a `/data` directory with `comic/` and `not-comic/` folders.
+
+### 3. Run the Training Notebook
+Open `comic-detector.ipynb` in Jupyter Notebook or an equivalent environment to retrain the model or evaluate it.
+
+### 4. Format Images (Optional)
+If your dataset images are not in a consistent format, use `image-format-generalizer.ipynb` to preprocess them.
+
+## Model Usage
+
+The fine-tuned model can be deployed directly via the Hugging Face Inference API. Once uploaded, the model can be used to classify whether an image contains Comic Sans font.
+
+Example API usage (replace `your-username/comic-sans-detector` with your repository name):
+```python
+from transformers import pipeline
+
+classifier = pipeline("image-classification", model="your-username/comic-sans-detector")
+result = classifier("path/to/image.jpg")
+print(result)
+```
+
+## Fine-Tuning Process
+
+This model was fine-tuned on a previously fine-tuned font classification model, which itself was based on the ResNet-18 foundation model. The fine-tuning process was conducted using a custom dataset with two classes: `comic` and `not-comic`.
+
+## Acknowledgments
+
+This project is based on the original font identifier repository by [gaborcselle](https://huggingface.co/gaborcselle/font-identifier).
+
+## License
+
+Include your preferred license here (e.g., MIT, Apache 2.0, etc.).

.ipynb_checkpoints/image-format-generalizer-checkpoint.ipynb

@@ -0,0 +1,84 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d3fca531-0f68-4951-b168-db8ad2d25971",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from PIL import Image, UnidentifiedImageError\n",
+    "import os\n",
+    "import pillow_avif  # AVIF support for Pillow\n",
+    "\n",
+    "# Define paths to folders\n",
+    "data_dir = \"./data\"\n",
+    "folders = [\"comic\", \"not-comic\"]  # Both folders to process\n",
+    "output_format = \"png\"  # Target image format\n",
+    "\n",
+    "# Function to clean, convert, and rename images\n",
+    "def process_images(data_dir, folders, output_format):\n",
+    "    for folder in folders:\n",
+    "        folder_path = os.path.join(data_dir, folder)\n",
+    "        print(f\"Processing folder: {folder_path}\")\n",
+    "\n",
+    "        # Ensure the folder exists\n",
+    "        if not os.path.exists(folder_path):\n",
+    "            print(f\"Folder {folder_path} does not exist. Skipping.\")\n",
+    "            continue\n",
+    "\n",
+    "        # Sort files to preserve order and avoid overwriting\n",
+    "        image_count = 1  # Start numbering images\n",
+    "        for filename in sorted(os.listdir(folder_path)):\n",
+    "            file_path = os.path.join(folder_path, filename)\n",
+    "            if \"test_sample.png\" in filename:  # Ignore test_sample.png\n",
+    "                continue\n",
+    "\n",
+    "            try:\n",
+    "                # Open and convert image (supports AVIF and others)\n",
+    "                with Image.open(file_path) as img:\n",
+    "                    img = img.convert(\"RGB\")  # Ensure compatible format\n",
+    "                    new_filename = f\"{str(image_count).zfill(4)}.{output_format}\"\n",
+    "                    new_file_path = os.path.join(folder_path, new_filename)\n",
+    "\n",
+    "                    # Save as new file\n",
+    "                    img.save(new_file_path, format=output_format.upper())\n",
+    "                    print(f\"Converted: {filename} -> {new_filename}\")\n",
+    "\n",
+    "                    # Remove old file if different\n",
+    "                    if file_path != new_file_path:\n",
+    "                        os.remove(file_path)\n",
+    "\n",
+    "                    image_count += 1\n",
+    "\n",
+    "            except (UnidentifiedImageError, IOError) as e:\n",
+    "                print(f\"Invalid or unreadable file: {filename} ({e}). Deleting.\")\n",
+    "                os.remove(file_path)  # Delete invalid files\n",
+    "\n",
+    "# Run the processing function\n",
+    "process_images(data_dir, folders, output_format)\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.5"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

.jupyter/desktop-workspaces/default-37a8.jupyterlab-workspace

@@ -0,0 +1 @@
+{"data":{"layout-restorer:data":{"main":{"dock":{"type":"tab-area","currentIndex":2,"widgets":["notebook:image-format-generalizer.ipynb","notebook:comic-detector.ipynb"]},"current":"notebook:comic-detector.ipynb"},"down":{"size":0,"widgets":[]},"left":{"collapsed":false,"visible":true,"current":"filebrowser","widgets":["filebrowser","running-sessions","@jupyterlab/toc:plugin","extensionmanager.main-view"],"widgetStates":{"jp-running-sessions":{"sizes":[0.16666666666666666,0.16666666666666666,0.16666666666666666,0.16666666666666666,0.16666666666666666,0.16666666666666666],"expansionStates":[false,false,false,false,false,false]},"extensionmanager.main-view":{"sizes":[0.3333333333333333,0.3333333333333333,0.3333333333333333],"expansionStates":[false,false,false]}}},"right":{"collapsed":true,"visible":true,"widgets":["jp-property-inspector","debugger-sidebar"],"widgetStates":{"jp-debugger-sidebar":{"sizes":[0.2,0.2,0.2,0.2,0.2],"expansionStates":[false,false,false,false,false]}}},"relativeSizes":[0.13545601726929304,0.864543982730707,0],"top":{"simpleVisibility":true}},"docmanager:recents":{"opened":[{"path":"","contentType":"directory","root":"~/Desktop/comic-sans-detector"},{"path":"comic-detector.ipynb","contentType":"notebook","factory":"Notebook","root":"~/Desktop/comic-sans-detector"},{"path":"README.md","contentType":"file","factory":"Editor","root":"~/Desktop/comic-sans-detector"},{"path":"image-format-generalizer.ipynb","contentType":"notebook","factory":"Notebook","root":"~/Desktop/comic-sans-detector"},{"path":"Untitled.ipynb","contentType":"notebook","factory":"Notebook","root":"~/Desktop/comic-sans-detector"}],"closed":[{"path":"README.md","contentType":"file","factory":"Editor","root":"~/Desktop/comic-sans-detector"}]},"notebook:image-format-generalizer.ipynb":{"data":{"path":"image-format-generalizer.ipynb","factory":"Notebook"}},"notebook:comic-detector.ipynb":{"data":{"path":"comic-detector.ipynb","factory":"Notebook"}},"file-browser-filebrowser:cwd":{"path":""}},"metadata":{"id":"default"}}

README.md

@@ -0,0 +1,69 @@
+# Comic Sans Detector
+
+This repository contains a fine-tuned ResNet-18 model, specifically trained to detect whether an image contains Comic Sans font. It is a fine-tuning of a previously fine-tuned font classification model, based on the ResNet-18 foundation model.
+
+## Repository Contents
+
+- **`comic-detector.ipynb`**: A notebook that demonstrates the training and evaluation process for the Comic Sans detector using the fine-tuned ResNet-18 model.
+- **`image-format-generalizer.ipynb`**: A utility notebook for preparing and normalizing image datasets, ensuring consistent formatting across `/data` folders.
+
+## Dataset Structure (Not Included)
+
+The dataset used for training and evaluation should follow this structure:
+```
+/data
+├── comic/
+│   ├── image1.jpg
+│   ├── image2.png
+│   └── ...
+├── not-comic/
+│   ├── image1.jpg
+│   ├── image2.png
+│   └── ...
+```
+- **`comic/`**: Contains images labeled as featuring Comic Sans font.
+- **`not-comic/`**: Contains images labeled as not featuring Comic Sans font.
+
+⚠️ The dataset itself is not included in this repository. You must prepare and structure your dataset as described.
+
+## How to Use
+
+### 1. Clone the Repository
+```bash
+git clone https://huggingface.co/your-username/comic-sans-detector
+cd comic-sans-detector
+```
+
+### 2. Prepare the Dataset
+Ensure your dataset is properly structured under a `/data` directory with `comic/` and `not-comic/` folders.
+
+### 3. Run the Training Notebook
+Open `comic-detector.ipynb` in Jupyter Notebook or an equivalent environment to retrain the model or evaluate it.
+
+### 4. Format Images (Optional)
+If your dataset images are not in a consistent format, use `image-format-generalizer.ipynb` to preprocess them.
+
+## Model Usage
+
+The fine-tuned model can be deployed directly via the Hugging Face Inference API. Once uploaded, the model can be used to classify whether an image contains Comic Sans font.
+
+Example API usage (replace `your-username/comic-sans-detector` with your repository name):
+```python
+from transformers import pipeline
+
+classifier = pipeline("image-classification", model="your-username/comic-sans-detector")
+result = classifier("path/to/image.jpg")
+print(result)
+```
+
+## Fine-Tuning Process
+
+This model was fine-tuned on a previously fine-tuned font classification model, which itself was based on the ResNet-18 foundation model. The fine-tuning process was conducted using a custom dataset with two classes: `comic` and `not-comic`.
+
+## Acknowledgments
+
+This project is based on the original font identifier repository by [gaborcselle](https://huggingface.co/gaborcselle/font-identifier).
+
+## License
+
+Include your preferred license here (e.g., MIT, Apache 2.0, etc.).

config.json

@@ -0,0 +1,143 @@
+{
+  "_name_or_path": "gaborcselle/font-identifier",
+  "architectures": [
+    "ResNetForImageClassification"
+  ],
+  "depths": [
+    2,
+    2,
+    2,
+    2
+  ],
+  "downsample_in_bottleneck": false,
+  "downsample_in_first_stage": false,
+  "embedding_size": 64,
+  "hidden_act": "relu",
+  "hidden_sizes": [
+    64,
+    128,
+    256,
+    512
+  ],
+  "id2label": {
+    "0": "Agbalumo-Regular",
+    "1": "AlfaSlabOne-Regular",
+    "2": "ArchitectsDaughter-Regular",
+    "3": "Arial",
+    "4": "Arial Black",
+    "5": "Arial Bold",
+    "6": "Arial Bold Italic",
+    "7": "Avenir",
+    "8": "Bangers-Regular",
+    "9": "BlackOpsOne-Regular",
+    "10": "Courier",
+    "11": "Georgia",
+    "12": "Helvetica",
+    "13": "IBMPlexSans-Regular",
+    "14": "Inter-Regular",
+    "15": "KaushanScript-Regular",
+    "16": "Lato-Regular",
+    "17": "Lobster-Regular",
+    "18": "Lora-Regular",
+    "19": "Merriweather-Regular",
+    "20": "Niconne-Regular",
+    "21": "OpenSans-Bold",
+    "22": "OpenSans-Italic",
+    "23": "OpenSans-Light",
+    "24": "Pacifico-Regular",
+    "25": "PixelifySans-Regular",
+    "26": "PlayfairDisplay-Regular",
+    "27": "Poppins-Regular",
+    "28": "Rakkas-Regular",
+    "29": "Roboto-Regular",
+    "30": "RobotoMono-Regular",
+    "31": "RobotoSlab-Regular",
+    "32": "Rubik-Regular",
+    "33": "SpaceMono-Regular",
+    "34": "Tahoma",
+    "35": "Tahoma Bold",
+    "36": "Times New Roman",
+    "37": "Times New Roman Bold",
+    "38": "Times New Roman Bold Italic",
+    "39": "Times New Roman Italic",
+    "40": "TitilliumWeb-Regular",
+    "41": "Trebuchet MS",
+    "42": "Trebuchet MS Bold",
+    "43": "Trebuchet MS Bold Italic",
+    "44": "Trebuchet MS Italic",
+    "45": "Verdana",
+    "46": "Verdana Bold",
+    "47": "Verdana Bold Italic",
+    "48": "Verdana Italic"
+  },
+  "label2id": {
+    "Agbalumo-Regular": "0",
+    "AlfaSlabOne-Regular": "1",
+    "ArchitectsDaughter-Regular": "2",
+    "Arial": "3",
+    "Arial Black": "4",
+    "Arial Bold": "5",
+    "Arial Bold Italic": "6",
+    "Avenir": "7",
+    "Bangers-Regular": "8",
+    "BlackOpsOne-Regular": "9",
+    "Courier": "10",
+    "Georgia": "11",
+    "Helvetica": "12",
+    "IBMPlexSans-Regular": "13",
+    "Inter-Regular": "14",
+    "KaushanScript-Regular": "15",
+    "Lato-Regular": "16",
+    "Lobster-Regular": "17",
+    "Lora-Regular": "18",
+    "Merriweather-Regular": "19",
+    "Niconne-Regular": "20",
+    "OpenSans-Bold": "21",
+    "OpenSans-Italic": "22",
+    "OpenSans-Light": "23",
+    "Pacifico-Regular": "24",
+    "PixelifySans-Regular": "25",
+    "PlayfairDisplay-Regular": "26",
+    "Poppins-Regular": "27",
+    "Rakkas-Regular": "28",
+    "Roboto-Regular": "29",
+    "RobotoMono-Regular": "30",
+    "RobotoSlab-Regular": "31",
+    "Rubik-Regular": "32",
+    "SpaceMono-Regular": "33",
+    "Tahoma": "34",
+    "Tahoma Bold": "35",
+    "Times New Roman": "36",
+    "Times New Roman Bold": "37",
+    "Times New Roman Bold Italic": "38",
+    "Times New Roman Italic": "39",
+    "TitilliumWeb-Regular": "40",
+    "Trebuchet MS": "41",
+    "Trebuchet MS Bold": "42",
+    "Trebuchet MS Bold Italic": "43",
+    "Trebuchet MS Italic": "44",
+    "Verdana": "45",
+    "Verdana Bold": "46",
+    "Verdana Bold Italic": "47",
+    "Verdana Italic": "48"
+  },
+  "layer_type": "basic",
+  "model_type": "resnet",
+  "num_channels": 3,
+  "out_features": [
+    "stage4"
+  ],
+  "out_indices": [
+    4
+  ],
+  "problem_type": "single_label_classification",
+  "stage_names": [
+    "stem",
+    "stage1",
+    "stage2",
+    "stage3",
+    "stage4"
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.47.0"
+}

image-format-generalizer.ipynb

@@ -0,0 +1,84 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d3fca531-0f68-4951-b168-db8ad2d25971",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from PIL import Image, UnidentifiedImageError\n",
+    "import os\n",
+    "import pillow_avif  # AVIF support for Pillow\n",
+    "\n",
+    "# Define paths to folders\n",
+    "data_dir = \"./data\"\n",
+    "folders = [\"comic\", \"not-comic\"]  # Both folders to process\n",
+    "output_format = \"png\"  # Target image format\n",
+    "\n",
+    "# Function to clean, convert, and rename images\n",
+    "def process_images(data_dir, folders, output_format):\n",
+    "    for folder in folders:\n",
+    "        folder_path = os.path.join(data_dir, folder)\n",
+    "        print(f\"Processing folder: {folder_path}\")\n",
+    "\n",
+    "        # Ensure the folder exists\n",
+    "        if not os.path.exists(folder_path):\n",
+    "            print(f\"Folder {folder_path} does not exist. Skipping.\")\n",
+    "            continue\n",
+    "\n",
+    "        # Sort files to preserve order and avoid overwriting\n",
+    "        image_count = 1  # Start numbering images\n",
+    "        for filename in sorted(os.listdir(folder_path)):\n",
+    "            file_path = os.path.join(folder_path, filename)\n",
+    "            if \"test_sample.png\" in filename:  # Ignore test_sample.png\n",
+    "                continue\n",
+    "\n",
+    "            try:\n",
+    "                # Open and convert image (supports AVIF and others)\n",
+    "                with Image.open(file_path) as img:\n",
+    "                    img = img.convert(\"RGB\")  # Ensure compatible format\n",
+    "                    new_filename = f\"{str(image_count).zfill(4)}.{output_format}\"\n",
+    "                    new_file_path = os.path.join(folder_path, new_filename)\n",
+    "\n",
+    "                    # Save as new file\n",
+    "                    img.save(new_file_path, format=output_format.upper())\n",
+    "                    print(f\"Converted: {filename} -> {new_filename}\")\n",
+    "\n",
+    "                    # Remove old file if different\n",
+    "                    if file_path != new_file_path:\n",
+    "                        os.remove(file_path)\n",
+    "\n",
+    "                    image_count += 1\n",
+    "\n",
+    "            except (UnidentifiedImageError, IOError) as e:\n",
+    "                print(f\"Invalid or unreadable file: {filename} ({e}). Deleting.\")\n",
+    "                os.remove(file_path)  # Delete invalid files\n",
+    "\n",
+    "# Run the processing function\n",
+    "process_images(data_dir, folders, output_format)\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.5"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb04f92ff671597997d664f096dc4d23fc3883a89c0d04db44acfbad76c696da
+size 44764336

preprocessor_config.json

@@ -0,0 +1,27 @@
+{
+  "crop_size": {
+    "height": 224,
+    "width": 224
+  },
+  "do_center_crop": true,
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.485,
+    0.456,
+    0.406
+  ],
+  "image_processor_type": "CLIPImageProcessor",
+  "image_std": [
+    0.229,
+    0.224,
+    0.225
+  ],
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "shortest_edge": 224
+  }
+}