{
"cells": [
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "view-in-github"
},
"source": [
"<a href=\"https://colab.research.google.com/github/ivelin/donut_ui_refexp/blob/main/Inference_Playground_Donut_UI_RefExp_Gradio.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "x6dFfL0QUr8P",
"outputId": "58f3b497-f4e8-46bc-a40c-b564f6e14010"
},
"outputs": [],
"source": [
"#@title Check out source repo if not automatically available\n",
"# !git clone https://github.com/GuardianUI/ui-refexp\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "RQdzURjDWYco",
"outputId": "2628c536-780e-4544-8f37-33a7e79ee367"
},
"outputs": [],
"source": [
"# Go to hf space dir if not already there\n",
"# !cd ui-refexp/hf-space && \n",
"\n",
"!pip3 install -r requirements.txt"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# from PIL import Image, ImageDraw\n",
"# from transformers import DonutProcessor, VisionEncoderDecoderModel\n",
"\n",
"# pretrained_repo_name = 'ivelin/donut-refexp-click'\n",
"# pretrained_revision = 'main'\n",
"# # revision can be git commit hash, branch or tag\n",
"# # use 'main' for latest revision\n",
"# print(f\"Loading model checkpoint: {pretrained_repo_name}\")\n",
"\n",
"# proc = DonutProcessor.from_pretrained(\n",
"# pretrained_repo_name, revision=pretrained_revision, use_auth_token=\"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\")\n",
"# proc.image_processor.do_align_long_axis = False\n",
"# proc.image_processor.do_resize = False\n",
"# proc.image_processor.do_thumbnail = False\n",
"# proc.image_processor.do_pad = False\n",
"# proc.image_processor.do_rescale = False\n"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/gitpod/.pyenv/versions/3.8.16/lib/python3.8/site-packages/tqdm/auto.py:22: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
" from .autonotebook import tqdm as notebook_tqdm\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Loading model checkpoint: ivelin/donut-refexp-click\n",
"processor image size: {'height': 1280, 'width': 960}\n",
"Running on local URL: http://127.0.0.1:7860\n",
"Running on public URL: https://f2beb057-2b06-4a52.gradio.live\n",
"\n",
"This share link expires in 72 hours. For free permanent hosting and GPU upgrades (NEW!), check out Spaces: https://huggingface.co/spaces\n"
]
},
{
"data": {
"text/html": [
"<div><iframe src=\"https://f2beb057-2b06-4a52.gradio.live\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
],
"text/plain": [
"<IPython.core.display.HTML object>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"(image, prompt): <PIL.Image.Image image mode=RGB size=2719x980 at 0x7F8F12C6E3D0>, click on search button\n",
"predicted decoder sequence: <s_refexp><s_prompt> click on search button</s_prompt><s_target_center><s_x> 0.23</s_x><s_y> 0.33</s_y></s_target_center></s>\n",
"predicted decoder sequence before token2json: <s_prompt> click on search button</s_prompt><s_target_center><s_x> 0.23</s_x><s_y> 0.33</s_y></s_target_center>\n",
"predicted center_point with text coordinates: {'x': '0.23', 'y': '0.33'}\n",
"predicted center_point with float coordinates: {'x': 0.23, 'y': 0.33, 'decoder output sequence (before x,y adjustment)': '<s_prompt> click on search button</s_prompt><s_target_center><s_x> 0.23</s_x><s_y> 0.33</s_y></s_target_center>'}\n",
"input image size: (2719, 980)\n",
"processed prompt: <s_refexp><s_prompt>click on search button</s_prompt><s_target_center>\n",
"point={'x': 0.23, 'y': 0.33, 'decoder output sequence (before x,y adjustment)': '<s_prompt> click on search button</s_prompt><s_target_center><s_x> 0.23</s_x><s_y> 0.33</s_y></s_target_center>'}, input_image_size=(2719, 980), output_image_size=(960, 1280)\n",
">>> resized_width=960\n",
">>> resized_height=346\n",
"translated point={'x': 0.23, 'y': 1.2208092485549134, 'decoder output sequence (before x,y adjustment)': '<s_prompt> click on search button</s_prompt><s_target_center><s_x> 0.23</s_x><s_y> 0.33</s_y></s_target_center>'}, resized_image_size: (960, 346)\n",
"to image pixel values: x, y: (625, 1196)\n",
"(image, prompt): <PIL.Image.Image image mode=RGB size=2719x980 at 0x7F8F12C5C9D0>, click on search names\n",
"predicted decoder sequence: <s_refexp><s_prompt> click on search names</s_prompt><s_target_center><s_x> 0.5</s_x><s_y> 0.18</s_y></s_target_center></s>\n",
"predicted decoder sequence before token2json: <s_prompt> click on search names</s_prompt><s_target_center><s_x> 0.5</s_x><s_y> 0.18</s_y></s_target_center>\n",
"predicted center_point with text coordinates: {'x': '0.5', 'y': '0.18'}\n",
"predicted center_point with float coordinates: {'x': 0.5, 'y': 0.18, 'decoder output sequence (before x,y adjustment)': '<s_prompt> click on search names</s_prompt><s_target_center><s_x> 0.5</s_x><s_y> 0.18</s_y></s_target_center>'}\n",
"input image size: (2719, 980)\n",
"processed prompt: <s_refexp><s_prompt>click on search names</s_prompt><s_target_center>\n",
"point={'x': 0.5, 'y': 0.18, 'decoder output sequence (before x,y adjustment)': '<s_prompt> click on search names</s_prompt><s_target_center><s_x> 0.5</s_x><s_y> 0.18</s_y></s_target_center>'}, input_image_size=(2719, 980), output_image_size=(960, 1280)\n",
">>> resized_width=960\n",
">>> resized_height=346\n",
"translated point={'x': 0.5, 'y': 0.6658959537572254, 'decoder output sequence (before x,y adjustment)': '<s_prompt> click on search names</s_prompt><s_target_center><s_x> 0.5</s_x><s_y> 0.18</s_y></s_target_center>'}, resized_image_size: (960, 346)\n",
"to image pixel values: x, y: (1359, 652)\n"
]
}
],
"source": [
"import app\n",
"\n",
"# img = Image.open('val-image-4.jpg')\n",
"# print(img.size)\n",
"# display(img)\n",
"# out_size = (proc.image_processor.size['width'],\n",
"# proc.image_processor.size['height'])\n",
"# oimg = app.prepare_image_for_encoder(img, output_image_size=out_size)\n",
"# print(oimg.size)\n",
"# display(oimg)\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# import transformers\n",
"\n",
"# turn off normalization so we can see the image\n",
"# otherwise its tiny [0..1] float values that all look like the color black(0)\n",
"# proc.image_processor.do_normalize = False\n",
"\n",
"# npimg = proc.image_processor.preprocess(oimg)\n",
"# pimg = transformers.image_transforms.to_pil_image(npimg['pixel_values'][0])\n",
"# pimg.save('tmp.png')\n",
"# display(pimg)\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"accelerator": "GPU",
"colab": {
"include_colab_link": true,
"provenance": []
},
"gpuClass": "standard",
"kernelspec": {
"display_name": "Python 3",
"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": "3.8.16"
},
"vscode": {
"interpreter": {
"hash": "9ac03a0a6051494cc606d484d27d20fce22fb7b4d169f583271e11d5ba46a56e"
}
}
},
"nbformat": 4,
"nbformat_minor": 0
}