{
"cells": [
{
"cell_type": "code",
"execution_count": 200,
"metadata": {},
"outputs": [],
"source": [
"import glob\n",
"from PIL import Image\n",
"import pillow_heif\n",
"import pandas as pd\n",
"import numpy as np\n",
"from torchvision.utils import save_image\n",
"import torchvision.transforms.functional as fn\n",
"import torchvision.transforms as transforms\n",
"import matplotlib.pyplot as plt"
]
},
{
"cell_type": "code",
"execution_count": 201,
"metadata": {},
"outputs": [],
"source": [
"# generates new dataset by pasting beetles and non-beetles in the same picture. \n",
"#returns array of new images and coordinates\n",
"#TODO: make sure that the beetles and non-beetles don't overlap with each other\n",
"\n",
"def generate_data(bg: list[Image], beetle_set: list[Image], num_beetles_arr: list[int], \n",
" non_beetle_set: list[Image], num_non_beetles_arr: list[int], overlap: bool):\n",
" set_size = len(beetle_set)\n",
" set_non_size = len(non_beetle_set)\n",
" set_bg_size = len(bg)\n",
" sim_arr, coords_arr = [],[]\n",
"\n",
" for i, num_beetles in enumerate(num_beetles_arr):\n",
" num_non_beetles = num_non_beetles_arr[i]\n",
" bg_id = np.random.randint(0, set_bg_size)\n",
" bg_temp = bg[bg_id].copy()\n",
" width, height = bg_temp.size\n",
" beetle_coords = []\n",
" non_beetle_coords = []\n",
" for j in range(num_non_beetles):\n",
" # get random non_beetle image\n",
" non_beetle_id = np.random.randint(0, set_non_size)\n",
" non_beetle_img = non_beetle_set[non_beetle_id]\n",
" \n",
" # get random non_beetle rotation\n",
" angle = np.random.randint(0, 360)\n",
" non_beetle_img = non_beetle_img.rotate(angle, resample=Image.BICUBIC, expand=1)\n",
" \n",
" non_beetle_width, non_beetle_height = non_beetle_img.size\n",
" \n",
" # get random x,y coords to paste non-beetle\n",
" x = np.random.randint(0, width - non_beetle_width)\n",
" y = np.random.randint(0, height - non_beetle_height)\n",
" \n",
" if not overlap:\n",
" is_overlapping = True\n",
" while (is_overlapping):\n",
" is_overlapping = False\n",
" for k in range(j):\n",
" _, x_, y_, width_, height_, _ = non_beetle_coords[-(k+1)]\n",
" if (x_ - width_/2) < x and (x_ + width_/2) > x and (y_ - height_/2) < y and (y_ + height_/2) > y:\n",
" is_overlapping = True\n",
" x = np.random.randint(0, width - non_beetle_width)\n",
" y = np.random.randint(0, height - non_beetle_height)\n",
" break\n",
"\n",
" bg_temp.paste(non_beetle_img, box=(x,y), mask=non_beetle_img)\n",
" \n",
" #centers x and y for YOLOv5 PyTorch label\n",
" x += non_beetle_width/2\n",
" y += non_beetle_width/2\n",
" non_beetle_coords.append((non_beetle_id, x, y, non_beetle_width, non_beetle_height, angle))\n",
"\n",
" \n",
" for j in range(num_beetles):\n",
" # get random beetle image\n",
" beetle_id = np.random.randint(0, set_size)\n",
" beetle_img = beetle_set[beetle_id]\n",
" # get random beetle rotation\n",
" angle = np.random.randint(0, 360)\n",
" beetle_img = beetle_img.rotate(angle, resample=Image.BICUBIC, expand=1)\n",
" \n",
" #randomly resize beetle to be smaller as they are much bigger on image\n",
" \n",
" beetle_width, beetle_height = beetle_img.size\n",
" beetle_max = np.max([beetle_width, beetle_height])\n",
" factor = np.random.uniform((height/(5*9))/beetle_max, (height/(4*9))/beetle_max)\n",
" beetle_img = beetle_img.resize((int(factor * beetle_width), int(factor * beetle_height)), resample=Image.BICUBIC)\n",
" beetle_width, beetle_height = beetle_img.size\n",
" \n",
" # get random x,y coords to paste beetle\n",
" x = np.random.randint(0, width - beetle_width)\n",
" y = np.random.randint(0, height - beetle_height)\n",
" \n",
" if not overlap:\n",
" is_overlapping = True\n",
" while (is_overlapping):\n",
" is_overlapping = False\n",
" for k in range(j+num_non_beetles):\n",
" if (k < num_non_beetles):\n",
" _, x_, y_, width_, height_, _ = non_beetle_coords[-(k+1)]\n",
" else:\n",
" _, x_, y_, width_, height_, _ = beetle_coords[-(k+1-num_non_beetles)]\n",
" if (x_ - width_/2) < x and (x_ + width_/2) > x and (y_ - height_/2) < y and (y_ + height_/2) > y:\n",
" is_overlapping = True\n",
" x = np.random.randint(0, width - non_beetle_width)\n",
" y = np.random.randint(0, height - non_beetle_height)\n",
" break\n",
" \n",
"\n",
" bg_temp.paste(beetle_img, box=(x,y), mask=beetle_img)\n",
" \n",
" #centers x and y for YOLOv5 PyTorch label\n",
" x += beetle_width/2\n",
" y += beetle_height/2\n",
" beetle_coords.append((beetle_id, x, y, beetle_width, beetle_height, angle))\n",
" sim_arr.append(bg_temp)\n",
" coords_arr.append(beetle_coords)\n",
" \n",
" return sim_arr, coords_arr"
]
},
{
"cell_type": "code",
"execution_count": 202,
"metadata": {},
"outputs": [],
"source": [
"#???\n",
"def find_coeffs(source_coords, target_coords):\n",
" matrix = []\n",
" for s, t in zip(source_coords, target_coords):\n",
" matrix.append([t[0], t[1], 1, 0, 0, 0, -s[0]*t[0], -s[0]*t[1]])\n",
" matrix.append([0, 0, 0, t[0], t[1], 1, -s[1]*t[0], -s[1]*t[1]])\n",
" A = np.matrix(matrix, dtype=float)\n",
" B = np.array(source_coords).reshape(8)\n",
" res = np.dot(np.linalg.inv(A.T * A) * A.T, B)\n",
" return np.array(res).reshape(8)"
]
},
{
"cell_type": "code",
"execution_count": 203,
"metadata": {},
"outputs": [],
"source": [
"backgrounds = []\n",
"for file in glob.glob(r\"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/*\"):\n",
" if \"bg.png\" in file: #clean trap\n",
" bg = Image.open(file)\n",
" # map corners of trap to corners of image\n",
" coeffs = find_coeffs([(128,6), (1904,62), (2113,3137), (3,3228)], \n",
" [(0,0), (bg.size[0], 0), (bg.size[0], bg.size[1]), (0, bg.size[1])])\n",
" bg_flat = bg.transform(bg.size, Image.PERSPECTIVE, coeffs, Image.BICUBIC)\n",
" backgrounds.append(bg)\n",
" continue\n",
" bg = Image.open(file)\n",
" \n",
" backgrounds.append(bg);\n",
"\n",
"beetles = []\n",
"for file in glob.glob(r\"/raid/projects/akhot2/group-01-phys371-sp2023/crop/beetles/*\"):\n",
" b0 = Image.open(file)\n",
" beetles.append(b0);\n",
"non_beetles = []\n",
"for file in glob.glob(r\"/raid/projects/akhot2/group-01-phys371-sp2023/crop/non_beetles/*\"):\n",
" n_b0 = Image.open(file)\n",
" non_beetles.append(n_b0);"
]
},
{
"cell_type": "code",
"execution_count": 204,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"21\n",
"20\n",
"48\n"
]
}
],
"source": [
"print(len(backgrounds))\n",
"print(len(beetles))\n",
"print(len(non_beetles))"
]
},
{
"cell_type": "code",
"execution_count": 188,
"metadata": {},
"outputs": [],
"source": [
"beetle_counts = np.random.randint(0, 8, size=1250)\n",
"non_beetle_counts = np.random.randint(0, 15, size=1250)\n",
"overlap = True\n",
"sim_img_arr, coords_arr = generate_data(backgrounds, beetles, beetle_counts, non_beetles, non_beetle_counts, overlap)"
]
},
{
"cell_type": "code",
"execution_count": 189,
"metadata": {},
"outputs": [],
"source": [
"#exports images and coordinates in YOLOv5 PyTorch format\n",
"def export(img_arr, coords_arr):\n",
" s = \"\"\n",
" image_set = \"train\"\n",
" change = len(img_arr)*0.8\n",
" for i in range(len(img_arr)):\n",
" s = \"\"\n",
" img = img_arr[i]\n",
" for coord in coords_arr[i]:\n",
" c, x, y, w, h, a = coord\n",
" s += f\"{0} {x/img.size[0]} {y/img.size[1]} {w/img.size[0]} {h/img.size[1]}\\n\"\n",
" print(i)\n",
" if i > change:\n",
" image_set = \"test\"\n",
" with open(\"data/\" +image_set+ f\"/labels/sim{i}.txt\", \"w\") as f:\n",
" f.write(s)\n",
" img.save(\"data/\" +image_set+ f\"/images/sim{i}.png\")"
]
},
{
"cell_type": "code",
"execution_count": 190,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"0\n",
"1\n",
"2\n",
"3\n",
"4\n",
"5\n",
"6\n",
"7\n",
"8\n",
"9\n"
]
}
],
"source": [
"export(sim_img_arr, coords_arr)"
]
},
{
"cell_type": "code",
"execution_count": 20,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"3.4648"
]
},
"execution_count": 20,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"beetle_counts.mean()"
]
},
{
"cell_type": "code",
"execution_count": 22,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"2.528"
]
},
"execution_count": 22,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"non_beetle_counts.mean()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"format\n",
"0 x y width height (each value normalized to size of image)\n",
"\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": 49,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"RGBA\n",
"(2, 4)\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAMIAAAD8CAYAAAAlkXvsAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAAAKeklEQVR4nO3dT4xdZR3G8e9jKUgEI0UhBIlU0g0aU2pTMRCiiX9qN9WFCS60C5ISA4kuXFRJtC4h0YUhkpRAqIlCSJTIBrUhGDZGWrGWllLaKtFC08YYA25U6s/FeSe9aafl/pk5c+/M95O8OXfeOffOmZM+vWdOO++TqkJa6d611AcgTQODIGEQJMAgSIBBkACDIAFLEIQkm5McSXIsyY6+v740n/T57whJVgGvAp8FTgB7ga9U1cu9HYQ0j77fETYBx6rqz1X1H+AJYGvPxyCd55Kev971wN8GPj4BfOLcnZJsB7a3Dz/ew3FpGaqqDLtv30GY78DOuzarql3ALoAk/h8QLbq+L41OADcMfPxB4I2ej0E6T99B2AusS7I2yaXAncDTPR+DdJ5eL42q6u0k9wK/BlYBj1bVoT6PQZpPr7dPx+HPCBrXKD8s+y/LEgZBAgyCBBgECTAIEmAQJMAgSIBBkACDIAEGQQIMggQYBAkwCBJgECTAIEiAQZAAgyABBkECDIIEGAQJMAgSYBAkwCBIgEGQAIMgARMGIclrSV5Ksj/Jvja3JsmeJEfb9qqB/b/dmnKOJPn8pAcvLZSFeEf4dFWtr6qN7eMdwLNVtQ54tn1MkpvpFv39CLAZ+HFr0JGW3GJcGm0FdrfHu4EvDsw/UVX/rqq/AMfoGnSkJTdpEAr4TZI/tJYbgGur6iRA217T5udry7l+vhdNsj3JvrnLLWmxTbos/G1V9UaSa4A9SV65yL5DteWAjTnq30TvCFX1RtueBp6iu9Q5leQ6gLY93Xa3LUdTa+wgJHlPkivnHgOfAw7SNeBsa7ttA37ZHj8N3JnksiRrgXXAC+N+fWkhTXJpdC3wVJK51/lZVf0qyV7gySR3AX8FvgxQVYeSPAm8DLwN3FNVZyY6emmB2JijZcvGHGlEBkHCIEiAQZAAgyABBkECDIIEGAQJMAgSYBAkwCBIgEGQAIMgAQZBAgyCBBgECTAIEmAQJMAgSIBBkACDIAEGQQIMggQYBAkYIghJHk1yOsnBgbmRy0CSfLyVihxL8qO0JfKkaTDMO8JjdMUeg8YpA3kI2E635um6eV5TWjLvGISqeh74xznTI5WBtFWx31tVv6tujcmfDDxHWnLj/owwahnI9e3xufPSVJi0KORcFyoDGbokBLrGHLrLKKkX474jjFoGcqI9Pnd+XlW1q6o2DhQUSotq3CCMVAbSLp/eSnJru1v0tYHnSEuvqi46gMeBk8B/6f5mvwu4mu5u0dG2XTOw/33AceAI8IWB+Y10jTrHgQdp3QxDfP1yOMYZw/z5mhsWhWjZsihEGpFBkDAIEmAQJMAgSIBBkACDIAEGQQIMggQYBAkwCBJgECTAIEiAQZAAgyABBkECDIIEGAQJMAgSYBAkwCBIgEGQAIMgAQZBAgyCBIzfmLMzyetJ9rexZeBzNuZo9gyx9ugdwAbg4MDcTuBb8+x7M/An4DJgLd06p6va514APkm3RPwzDKyL6tqnjsUYo6x9Om5jzoVsxcYczaBJfka4N8mBduk0Vya4II05SbYn2Zdk3wTHJw1t3CA8BNwErKdbMv4HbX5BGnMsClHfxgpCVZ2qqjNV9T/gYWBT+9SCNOZIfRsrCHO1Uc2X6ApAwMYczah3LBNM8jjwKeD9SU4A3wM+lWQ93eXNa8DdAFV1KMmTwMvA28A9VXWmvdTX6TqbL6e7a/TMAn4f0kRszNGyZWOONCKDIGEQJMAgSIBBkACDIAEGQQIMggQYBAkwCBJgECTAIEiAQZAAgyABBkECDIIEGAQJMAgSYBAkwCBIgEGQAIMgAQZBAgyCBBgECRiuMeeGJM8lOZzkUJJvtPk1SfYkOdq2Vw08x9YczZYhGmuuAza0x1cCr9I14zwA7GjzO4D7F6M1hyloXnHM5ljoxpyTVfVie/wWcJiu5GMrsLvttpuzDThbsTVHM2aknxGS3AjcAvweuLYt907bXtN2m7g1x8Yc9e0dl4Wfk+QK4OfAN6vqzYtc3k/cmlNVu4Bd7evOu4+0kIZ6R0iymi4EP62qX7TpU3OFIW17us3bmqOZM8xdowCPAIer6ocDn3oa2NYeb+NsA46tOZo9Q9y1uZ3uEuYAsL+NLcDVwLPA0bZdM/Cc++juFh1h4M4QsJGuZuo48CCtqMS7Ro7FGKPcNbIxR8uWjTnSiAyChEGQAIMgAQZBAgyCBBgECTAIEmAQJMAgSIBBkACDIAEGQQIMggQYBAkwCBJgECTAIEiAQZAAgyABBkECDIIEGAQJMAgSMFlRyM4kryfZ38aWgedYFKLZMkFRyE7gW/Psb1GIYypGX0UhF7IVi0I0YyYpCgG4N8mBJI8OdKhZFKKZM3QQzi0KAR4CbgLWAyeBH8ztOs/T6yLz509W7aqqjVW1cdjjkyYxdlFIVZ2qqjNV9T/gYWBT292iEM2csYtC5tpymi/R9R6ARSGaQcN0qN0GfBV4Kcn+Nvcd4CtJ1tNd3rwG3A1QVYeSPAm8DLwN3FNVZ9rzvg48BlxOd9fomYX4JqRJWRSiZcuiEGlEBkHCIEiAQZAAgyABBkECDIIEGAQJMAgSYBAkwCBIgEGQAIMgAQZBAgyCBBgECTAIEmAQJMAgSIBBkACDIAEGQQIMggQYBAkwCBIw3Nqn707yQpI/tcac77f5NUn2JDnatlcNPMfGHM2WIRprAlzRHq+m60a4FXgA2NHmdwD325jjmKax0I05VVX/ah+ubqPomnF2t/ndnG2/2YqNOZoxw/YjrGorYZ8G9lTV74Fr21LvtO01bXcbczRzhgpCKwRZT1fusSnJRy+yu405mjkj3TWqqn8CvwU2A6fmykLa9nTbzcYczZxh7hp9IMn72uPLgc8Ar9A142xru23jbPuNjTmaPUPctfkY8EfgAF091Hfb/NXAs8DRtl0z8Jz76O4WHWHgzhCwsb3GceBBWlGJd40cizFGuWtkY46WLRtzpBEZBAmDIAEGQQIMggQYBAkwCBJgECTAIEiAQZAAuGSpD2AI/6L7P0uC9wN/X+qDmBLvdC4+NMqLzUIQjvh7CZ0k+zwXnYU+F14aSRgECZiNIOxa6gOYIp6Lsxb0XEz97yNIfZiFdwRp0RkEiSkOQpLNbcnIY0l2LPXxLIYkjyY5neTgwNyKXEozyQ1JnktyuC0t+o0238/5GOUXnPsawCq6X/D/MHAp3RKSNy/1cS3C93kHsAE4ODDXy1Ka0zaA64AN7fGVwKvte+7lfEzrO8Im4FhV/bmq/gM8QbeU5LJSVc8D/zhneisrcCnNqjpZVS+2x28Bh+lWQuzlfExrEC60bORKsGhLac6KJDcCt9AtON3L+ZjWIAy9POQKMvFSmrMgyRXAz4FvVtWbF9t1nrmxz8e0BuFCy0auBCt2Kc0kq+lC8NOq+kWb7uV8TGsQ9gLrkqxNcilwJ91SkivBilxKsx37I8DhqvrhwKf6OR9LfbfgIncRttDdOTgO3LfUx7NI3+PjwEngv3R/k91FT0tpTtsAbqe7hDkA7G9jS1/nw/9iITG9l0ZSrwyChEGQAIMgAQZBAgyCBBgECYD/A79FAQx8InyaAAAAAElFTkSuQmCC",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"from torchvision.utils import draw_bounding_boxes\n",
"import torch\n",
"import torchvision\n",
"transform = transforms.Compose([transforms.ToTensor()])\n",
"for i in range(len(sim_img_arr)):\n",
" print(sim_img_arr[i].mode)\n",
" \n",
" tensor = transform(sim_img_arr[i].convert(\"RGB\")).type(torch.uint8)\n",
" coords_list = np.zeros((len(coords_arr[i]), 4))\n",
" j=0\n",
" for k in coords_arr[i]:\n",
" c, x, y, w, h, a = k\n",
" coords_list[j, 0] = x\n",
" coords_list[j, 1] = y\n",
" coords_list[j, 2] = int(x+w)\n",
" coords_list[j, 3] = int(y+h)\n",
" j+=1\n",
" print(coords_list.shape)\n",
" draw_bounding_boxes(tensor, torch.Tensor(coords_list))\n",
" plt.imshow(tensor.permute(1,2,0)[:,:,0:3])\n",
" plt.show()\n",
" break\n",
" \n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"First:\n",
" - created 1250 images on a clean background variable 1 to 5 of the same beetle, model detected well on test set of 100 images (val/exp or exp2 or exp3) (proof of concept)\n",
" \n",
" fixed labels, making sure it's in the center\n",
" \n",
" - created 1250 images on a clean background variable 0 to 5 beetles of 6 different types of beetles, performing nearly 100%\n",
" https://wandb.ai/akhot2/YOLOv5/runs/294etqct?workspace=user-fresleven\n",
" https://wandb.ai/akhot2/YOLOv5/runs/i0ykd1a6?workspace=user-fresleven\n",
" \n",
"TODO:\n",
" - (fixed autocropping during rotation and added option for not overlapping and multiple backgrounds and non beetles)\n",
" - created 1250 images on a clean background variable 0 to 7 beetles of 6 different images and 0 to 5 non-beetles 7 different images, performing \n",
" - created 1250 images on a clean and dirty background variables 0 to 15 beetles of 15 different images and 0 to 10 non-beetle of 10 different images, overlap True\n",
" \n",
" - created 1250 images on a clean and dirty background variables 0 to 10 beetles of 15 different images and 0 to 10 non-beetle of 10 different images, overlap False\n",
" \n",
" - perform auto cropping on arduino and include "
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"4"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"len(backgrounds)"
]
},
{
"cell_type": "code",
"execution_count": 196,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Using cache found in /home/akhot2/.cache/torch/hub/ultralytics_yolov5_master\n",
"YOLOv5 🚀 2023-3-31 Python-3.9.12 torch-1.11.0 CUDA:0 (A100-SXM4-40GB, 40537MiB)\n",
"\n",
"Fusing layers... \n",
"Model summary: 212 layers, 20852934 parameters, 0 gradients, 47.9 GFLOPs\n",
"Adding AutoShape... \n"
]
},
{
"data": {
"text/plain": [
"AutoShape(\n",
" (model): DetectMultiBackend(\n",
" (model): DetectionModel(\n",
" (model): Sequential(\n",
" (0): Conv(\n",
" (conv): Conv2d(3, 48, kernel_size=(6, 6), stride=(2, 2), padding=(2, 2))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (1): Conv(\n",
" (conv): Conv2d(48, 96, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (2): C3(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(96, 48, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(96, 48, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv3): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): Sequential(\n",
" (0): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(48, 48, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (1): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(48, 48, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" )\n",
" )\n",
" (3): Conv(\n",
" (conv): Conv2d(96, 192, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (4): C3(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv3): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): Sequential(\n",
" (0): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (1): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (2): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (3): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" )\n",
" )\n",
" (5): Conv(\n",
" (conv): Conv2d(192, 384, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (6): C3(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(384, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(384, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv3): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): Sequential(\n",
" (0): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (1): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (2): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (3): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (4): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (5): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" )\n",
" )\n",
" (7): Conv(\n",
" (conv): Conv2d(384, 768, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (8): C3(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(768, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(768, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv3): Conv(\n",
" (conv): Conv2d(768, 768, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): Sequential(\n",
" (0): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (1): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" )\n",
" )\n",
" (9): SPPF(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(768, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(1536, 768, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): MaxPool2d(kernel_size=5, stride=1, padding=2, dilation=1, ceil_mode=False)\n",
" )\n",
" (10): Conv(\n",
" (conv): Conv2d(768, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (11): Upsample(scale_factor=2.0, mode=nearest)\n",
" (12): Concat()\n",
" (13): C3(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(768, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(768, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv3): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): Sequential(\n",
" (0): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (1): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" )\n",
" )\n",
" (14): Conv(\n",
" (conv): Conv2d(384, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (15): Upsample(scale_factor=2.0, mode=nearest)\n",
" (16): Concat()\n",
" (17): C3(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv3): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): Sequential(\n",
" (0): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (1): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" )\n",
" )\n",
" (18): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (19): Concat()\n",
" (20): C3(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(384, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(384, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv3): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): Sequential(\n",
" (0): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (1): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" )\n",
" )\n",
" (21): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (22): Concat()\n",
" (23): C3(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(768, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(768, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv3): Conv(\n",
" (conv): Conv2d(768, 768, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (m): Sequential(\n",
" (0): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" (1): Bottleneck(\n",
" (cv1): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" (cv2): Conv(\n",
" (conv): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (act): SiLU(inplace=True)\n",
" )\n",
" )\n",
" )\n",
" )\n",
" (24): Detect(\n",
" (m): ModuleList(\n",
" (0): Conv2d(192, 18, kernel_size=(1, 1), stride=(1, 1))\n",
" (1): Conv2d(384, 18, kernel_size=(1, 1), stride=(1, 1))\n",
" (2): Conv2d(768, 18, kernel_size=(1, 1), stride=(1, 1))\n",
" )\n",
" )\n",
" )\n",
" )\n",
" )\n",
")"
]
},
"execution_count": 196,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import torch\n",
"model = torch.hub.load('ultralytics/yolov5', 'custom', '/raid/projects/akhot2/group-01-phys371-sp2023/yolov5_model/runs/train/20beetle_40-non_20dirt_bkg_overlap/weights/best.pt')\n",
"model.eval()"
]
},
{
"cell_type": "code",
"execution_count": 199,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"YOLOv5 <class 'models.common.Detections'> instance\n",
"image 1/1: 2746x1610 3 beetless\n",
"Speed: 125.2ms pre-process, 13.4ms inference, 1.1ms NMS per image at shape (1, 3, 640, 384)"
]
},
"execution_count": 199,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model('/raid/projects/akhot2/group-01-phys371-sp2023/crop/data/train/images/sim1.png')"
]
},
{
"cell_type": "code",
"execution_count": 153,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real18.jpg\n",
"(1744, 2778)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real16.jpg\n",
"(1894, 3155)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real7.jpg\n",
"(1837, 2978)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real8.jpg\n",
"(1680, 2791)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real10.jpg\n",
"(1880, 3101)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real13.jpg\n",
"(1572, 2582)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real15.jpg\n",
"(1486, 2422)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real6.jpg\n",
"(1635, 2699)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/F28_3AUGUST2022_cropped.jpg\n",
"(1850, 2765)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/bg.png\n",
"(2114, 3236)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real17.jpg\n",
"(1769, 3095)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real1.jpg\n",
"(1751, 2631)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real20.jpg\n",
"(1521, 2306)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real4.jpg\n",
"(1640, 2289)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real3.jpg\n",
"(1610, 2746)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real19.jpg\n",
"(1662, 2767)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real9.jpg\n",
"(1627, 2155)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real14.jpg\n",
"(1548, 2187)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real5.jpg\n",
"(1940, 2713)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real12.jpg\n",
"(1494, 2695)\n",
"/raid/projects/akhot2/group-01-phys371-sp2023/crop/cropped_imgs/real2.jpg\n",
"(1940, 2713)\n"
]
}
],
"source": [
"width = []\n",
"height = []\n",
"for file in glob.glob('/raid/projects/akhot2/group-01-phys371-sp2023/crop/data/train/sim0.png'):\n",
" print(file)\n",
" n_b0 = Image.open(file)\n",
" print(n_b0.size)\n",
" w, h = n_b0.size\n",
" width.append(w)\n",
" height.append(h)\n",
" #f = model(file)\n",
" #print(f)"
]
},
{
"cell_type": "code",
"execution_count": 76,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"78.57894736842105"
]
},
"execution_count": 76,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"np.mean(width)"
]
},
{
"cell_type": "code",
"execution_count": 77,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"92.47368421052632"
]
},
"execution_count": 77,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"np.mean(height)"
]
},
{
"cell_type": "code",
"execution_count": 154,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"1723.5238095238096"
]
},
"execution_count": 154,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"np.mean(width)"
]
},
{
"cell_type": "code",
"execution_count": 155,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"2704.9523809523807"
]
},
"execution_count": 155,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"np.mean(height)"
]
},
{
"cell_type": "code",
"execution_count": 93,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"94.5"
]
},
"execution_count": 93,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"width[8]*.7"
]
},
{
"cell_type": "code",
"execution_count": 94,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"54.599999999999994"
]
},
"execution_count": 94,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"height[8]*.7"
]
},
{
"cell_type": "code",
"execution_count": 156,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"[1744,\n",
" 1894,\n",
" 1837,\n",
" 1680,\n",
" 1880,\n",
" 1572,\n",
" 1486,\n",
" 1635,\n",
" 1850,\n",
" 2114,\n",
" 1769,\n",
" 1751,\n",
" 1521,\n",
" 1640,\n",
" 1610,\n",
" 1662,\n",
" 1627,\n",
" 1548,\n",
" 1940,\n",
" 1494,\n",
" 1940]"
]
},
"execution_count": 156,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"width"
]
},
{
"cell_type": "code",
"execution_count": 166,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"51.44444444444444\n",
"41.15555555555556\n",
"58.425925925925924\n",
"46.74074074074074\n",
"55.148148148148145\n",
"44.11851851851852\n",
"51.68518518518518\n",
"41.34814814814815\n",
"57.425925925925924\n",
"45.94074074074074\n",
"47.81481481481482\n",
"38.25185185185185\n",
"44.851851851851855\n",
"35.88148148148148\n",
"49.98148148148148\n",
"39.98518518518519\n",
"51.2037037037037\n",
"40.96296296296296\n",
"59.925925925925924\n",
"47.94074074074074\n",
"57.31481481481482\n",
"45.851851851851855\n",
"48.72222222222222\n",
"38.977777777777774\n",
"42.7037037037037\n",
"34.162962962962965\n",
"42.388888888888886\n",
"33.91111111111111\n",
"50.851851851851855\n",
"40.681481481481484\n",
"51.24074074074074\n",
"40.992592592592594\n",
"39.907407407407405\n",
"31.925925925925927\n",
"40.5\n",
"32.4\n",
"50.24074074074074\n",
"40.19259259259259\n",
"49.907407407407405\n",
"39.925925925925924\n",
"50.24074074074074\n",
"40.19259259259259\n"
]
}
],
"source": [
"for i in height:\n",
" print((i/(6*9)))\n",
" print((i/(7.5*9)))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"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": "3.10.9"
},
"vscode": {
"interpreter": {
"hash": "5501471bf458387d76cdb9f487659a1abd30d9114d99ffb68b67c5dd3feef292"
}
}
},
"nbformat": 4,
"nbformat_minor": 2
}