Release

config.json

@@ -0,0 +1,32 @@
+{
+  "_name_or_path": "shadowlilac/anime-image-quality-v2",
+  "architectures": [
+    "ViTForImageClassification"
+  ],
+  "attention_probs_dropout_prob": 0.0,
+  "encoder_stride": 16,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.0,
+  "hidden_size": 1536,
+  "id2label": {
+    "0": "hq",
+    "1": "lq"
+  },
+  "image_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4192,
+  "label2id": {
+    "hq": "0",
+    "lq": "1"
+  },
+  "layer_norm_eps": 1e-12,
+  "model_type": "vit",
+  "num_attention_heads": 16,
+  "num_channels": 3,
+  "num_hidden_layers": 48,
+  "patch_size": 64,
+  "problem_type": "single_label_classification",
+  "qkv_bias": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.33.2"
+}

inference.ipynb

@@ -0,0 +1,148 @@
+{
+  "nbformat": 4,
+  "nbformat_minor": 0,
+  "metadata": {
+    "colab": {
+      "provenance": [],
+      "collapsed_sections": [
+        "3xnrF3UB6ev0"
+      ],
+      "gpuType": "T4"
+    },
+    "kernelspec": {
+      "name": "python3",
+      "display_name": "Python 3"
+    },
+    "language_info": {
+      "name": "python"
+    },
+    "accelerator": "GPU"
+  },
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "source": [
+        "# Model Inference"
+      ],
+      "metadata": {
+        "id": "33C47swS80_1"
+      }
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "#@title Install Dependencies\n",
+        "!pip install transformers -q"
+      ],
+      "metadata": {
+        "cellView": "form",
+        "id": "noaoheUjvGbd"
+      },
+      "execution_count": 1,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "cellView": "form",
+        "id": "NZLqjuWEtCDy"
+      },
+      "outputs": [],
+      "source": [
+        "#@title Imports\n",
+        "import os\n",
+        "from transformers import pipeline\n",
+        "import shutil\n",
+        "from PIL import Image\n",
+        "import torch\n",
+        "pipe = pipeline(\"image-classification\", model=\"shadowlilac/aesthetic-shadow\", device=0)"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "#@title Inference\n",
+        "\n",
+        "# Input image file\n",
+        "single_image_file = \"image_1.png\" #@param {type:\"string\"}\n",
+        "\n",
+        "result = pipe(images=[single_image_file])\n",
+        "\n",
+        "prediction_single = result[0]\n",
+        "print(\"Prediction: \" + str(round([p for p in prediction_single if p['label'] == 'hq'][0]['score'], 2)) + \"% High Quality\")\n",
+        "Image.open(single_image_file)"
+      ],
+      "metadata": {
+        "cellView": "form",
+        "id": "r1R-L2r-0uo2"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "source": [
+        "# Batch Mode"
+      ],
+      "metadata": {
+        "id": "3xnrF3UB6ev0"
+      }
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "#@title Batch parameters\n",
+        "# Define the paths for the input folder and output folders\n",
+        "input_folder = \"input_folder\" #@param {type:\"string\"}\n",
+        "output_folder_hq = \"output_hq_folder\" #@param {type:\"string\"}\n",
+        "output_folder_lq = \"output_lq_folder\" #@param {type:\"string\"}\n",
+        "# Threshhold\n",
+        "batch_hq_threshold = 0.5 #@param {type:\"number\"}\n",
+        "# Define the batch size\n",
+        "batch_size = 8 #@param {type:\"number\"}"
+      ],
+      "metadata": {
+        "cellView": "form",
+        "id": "VlPgrJf4wpHo"
+      },
+      "execution_count": 4,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "#@title Execute Batch Job\n",
+        "\n",
+        "# List all image files in the input folder\n",
+        "image_files = [os.path.join(input_folder, f) for f in os.listdir(input_folder) if f.lower().endswith(('.png', '.jpg', '.jpeg'))]\n",
+        "\n",
+        "# Process images in batches\n",
+        "for i in range(0, len(image_files), batch_size):\n",
+        "    batch = image_files[i:i + batch_size]\n",
+        "\n",
+        "    # Perform classification for the batch\n",
+        "    results = pipe(images=batch)\n",
+        "\n",
+        "    for idx, result in enumerate(results):\n",
+        "        # Extract the prediction scores and labels\n",
+        "        predictions = result\n",
+        "        hq_score = [p for p in predictions if p['label'] == 'hq'][0]['score']\n",
+        "\n",
+        "        # Determine the destination folder based on the prediction and threshold\n",
+        "        destination_folder = output_folder_hq if hq_score >= batch_hq_threshold else output_folder_lq\n",
+        "\n",
+        "        # Copy the image to the appropriate folder\n",
+        "        shutil.copy(batch[idx], os.path.join(destination_folder, os.path.basename(batch[idx])))\n",
+        "\n",
+        "print(\"Classification and sorting complete.\")"
+      ],
+      "metadata": {
+        "cellView": "form",
+        "id": "RG01mcYf4DvK"
+      },
+      "execution_count": null,
+      "outputs": []
+    }
+  ]
+}

preprocessor_config.json

@@ -0,0 +1,22 @@
+{
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "ViTFeatureExtractor",
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 1024,
+    "width": 1024
+  }
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7eac1cb6aa06d1a82fa162e124bfbcd6aaaa47dcfbcb8d1a628618e3c1d6f581
+size 4365309073