Skip to content

Instantly share code, notes, and snippets.

@heyodai

heyodai/fix-makesense-annotations.ipynb Secret

Last active January 28, 2024 20:59
Show Gist options
  • Save heyodai/43548d252b6ad03f54e400dbdfc93778 to your computer and use it in GitHub Desktop.
Save heyodai/43548d252b6ad03f54e400dbdfc93778 to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
fix-makesense-annotations.ipynb
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Reference: https://github.com/SkalskiP/make-sense/issues/351\n",
"\n",
"[makesense.ai](https://www.makesense.ai/) is not exporting my annotations properly to YOLO format. I believe this a bug caused by some files having the same file names (e.g. `1.jpg`).\n",
"\n",
"Fortunately, the CSV export works as expected. This notebook transforms the CSV export to YOLO format."
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"import pandas as pd\n",
"import os\n",
"\n",
"BASE_FP = './dataset'\n",
"DATASET_FP = f'{BASE_FP}/v3'\n",
"CLASSES = ['thrip', 'healthy']"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"!rm -rf {DATASET_FP}"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"# Create dataset folders\n",
"os.makedirs(DATASET_FP, exist_ok=True)\n",
"os.makedirs(f\"{DATASET_FP}/images\", exist_ok=True)\n",
"os.makedirs(f\"{DATASET_FP}/labels\", exist_ok=True)\n",
"\n",
"# Create `classes.txt` file\n",
"with open(f\"{DATASET_FP}/classes.txt\", 'w') as f:\n",
" for c in CLASSES:\n",
" f.write(f\"{c}\\n\")\n",
" \n",
"# Create `dataset.yaml` file\n",
"data = f\"\"\"\n",
"path: {DATASET_FP} \n",
"train: {DATASET_FP}/images\n",
"\n",
"nc: {len(CLASSES)}\n",
"names: {CLASSES}\n",
"\"\"\"\n",
"with open(f\"{DATASET_FP}/dataset.yaml\", 'w') as f:\n",
" f.write(data)"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>label_name</th>\n",
" <th>bbox_x</th>\n",
" <th>bbox_y</th>\n",
" <th>bbox_width</th>\n",
" <th>bbox_height</th>\n",
" <th>image_name</th>\n",
" <th>image_width</th>\n",
" <th>image_height</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>thrip</td>\n",
" <td>0</td>\n",
" <td>323</td>\n",
" <td>1077</td>\n",
" <td>582</td>\n",
" <td>1.jpg</td>\n",
" <td>1080</td>\n",
" <td>1920</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>thrip</td>\n",
" <td>0</td>\n",
" <td>7</td>\n",
" <td>454</td>\n",
" <td>499</td>\n",
" <td>1.jpg</td>\n",
" <td>1080</td>\n",
" <td>1920</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>thrip</td>\n",
" <td>10</td>\n",
" <td>912</td>\n",
" <td>1063</td>\n",
" <td>994</td>\n",
" <td>1.jpg</td>\n",
" <td>1080</td>\n",
" <td>1920</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>thrip</td>\n",
" <td>7</td>\n",
" <td>151</td>\n",
" <td>523</td>\n",
" <td>1762</td>\n",
" <td>10.jpg</td>\n",
" <td>1080</td>\n",
" <td>1920</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>thrip</td>\n",
" <td>468</td>\n",
" <td>179</td>\n",
" <td>602</td>\n",
" <td>1734</td>\n",
" <td>10.jpg</td>\n",
" <td>1080</td>\n",
" <td>1920</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" label_name bbox_x bbox_y bbox_width bbox_height image_name image_width \\\n",
"0 thrip 0 323 1077 582 1.jpg 1080 \n",
"1 thrip 0 7 454 499 1.jpg 1080 \n",
"2 thrip 10 912 1063 994 1.jpg 1080 \n",
"3 thrip 7 151 523 1762 10.jpg 1080 \n",
"4 thrip 468 179 602 1734 10.jpg 1080 \n",
"\n",
" image_height \n",
"0 1920 \n",
"1 1920 \n",
"2 1920 \n",
"3 1920 \n",
"4 1920 "
]
},
"execution_count": 4,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"df = pd.read_csv('makesense.csv')\n",
"df.head()"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(176, 8)"
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"df.shape"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"last_match = None\n",
"file_count = 1\n",
"\n",
"for index, row in df.iterrows():\n",
" class_name = row['label_name']\n",
" file_name = row['image_name'].split('.')[0]\n",
" \n",
" input_fp = f\"{BASE_FP}/images/{class_name}/{file_name}.jpg\"\n",
" output_fp = f\"{DATASET_FP}/images/image{file_count}.jpg\"\n",
" \n",
" # Copy if not already copied\n",
" new_match = f\"{class_name}_{file_name}\"\n",
" if last_match == None or last_match != new_match:\n",
" os.system(f\"cp {input_fp} {output_fp}\")\n",
" file_count += 1\n",
" last_match = new_match\n",
" \n",
" # Create the label row\n",
" cid = CLASSES.index(class_name)\n",
" x = row['bbox_x']\n",
" y = row['bbox_y']\n",
" w = row['bbox_width']\n",
" h = row['bbox_height']\n",
" label_format = f\"{cid} {x} {y} {w} {h}\"\n",
" label_file = f\"{DATASET_FP}/labels/image{file_count}.txt\"\n",
" \n",
" # Append to the label file\n",
" with open(label_file, 'a') as f:\n",
" f.write(f\"{label_format}\\n\")\n",
" "
]
}
],
"metadata": {
"kernelspec": {
"display_name": ".venv",
"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": "3.11.6"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment