Merge pull request #7 from BU-Spark/project-setup

Project setup

documents/PoC/Mustafa/Mustafa_PoC.ipynb

@@ -0,0 +1,188 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "fc8d75bc-5e9e-4fe6-a683-ae07437eac38",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import json\n",
+    "import torch\n",
+    "import torch.nn as nn\n",
+    "from torchvision import transforms\n",
+    "from torch.utils.data import DataLoader\n",
+    "from PIL import Image\n",
+    "from yolov5 import YOLOv5"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "aa49ffdc-15ed-45ad-8e3b-770b727e404b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class CustomDataset(Dataset):\n",
+    "    def __init__(self, directory, transform=None):\n",
+    "        self.directory = directory\n",
+    "        self.transform = transform\n",
+    "        self.images = []\n",
+    "        self.annotations = []\n",
+    "\n",
+    "        for filename in os.listdir(directory):\n",
+    "            if filename.endswith('.jpeg'):\n",
+    "                img_path = os.path.join(directory, filename)\n",
+    "                ann_path = os.path.join(directory, filename.replace('.jpeg', '.json'))\n",
+    "                \n",
+    "                if os.path.exists(ann_path):\n",
+    "                    self.images.append(img_path)\n",
+    "                    self.annotations.append(ann_path)\n",
+    "\n",
+    "    def __len__(self):\n",
+    "        return len(self.images)\n",
+    "\n",
+    "    def __getitem__(self, idx):\n",
+    "        img_path = self.images[idx]\n",
+    "        ann_path = self.annotations[idx]\n",
+    "\n",
+    "        image = Image.open(img_path).convert('RGB')\n",
+    "        with open(ann_path, 'r') as f:\n",
+    "            annotation = json.load(f)\n",
+    "\n",
+    "        categories = [ann['category_id'] for ann in annotation['annotations']]\n",
+    "        bboxes = [ann['bbox'] for ann in annotation['annotations']]\n",
+    "\n",
+    "        image_width, image_height = image.size\n",
+    "        bboxes = [[(x + w / 2) / image_width, (y + h / 2) / image_height, w / image_width, h / image_height] \n",
+    "                  for x, y, w, h in bboxes]\n",
+    "\n",
+    "        target = {'boxes': torch.Tensor(bboxes), 'labels': torch.Tensor(categories).long()}\n",
+    "        \n",
+    "        if self.transform:\n",
+    "            image = self.transform(image)\n",
+    "\n",
+    "        return image, target\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "acb0de3a-d9fa-444e-ae5d-090f0b3a673e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "transform = transforms.Compose([\n",
+    "    transforms.Resize((640, 640)),\n",
+    "    transforms.ToTensor(),\n",
+    "])\n",
+    "\n",
+    "dataset = CustomDataset(directory=\"train\", transform=transform)\n",
+    "data_loader = DataLoader(dataset, batch_size=16, shuffle=True, num_workers=4)\n",
+    "\n",
+    "model = YOLOv5('yolov5s.pt', autoshape=True)\n",
+    "model.train()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "cec6bbdf-b14e-405c-87d5-4bcb0edab308",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Loss function\n",
+    "class YOLOv5Loss(nn.Module):\n",
+    "    def __init__(self):\n",
+    "        super(YOLOv5Loss, self).__init__()\n",
+    "        self.mse_loss = nn.MSELoss() # GIoU another option\n",
+    "        self.bce_loss = nn.BCEWithLogitsLoss()\n",
+    "        self.ce_loss = nn.CrossEntropyLoss()\n",
+    "\n",
+    "    def forward(self, predictions, targets):\n",
+    "        # Implement parsing of predictions and targets\n",
+    "        obj_preds, no_obj_preds, class_preds, box_preds = predictions\n",
+    "        obj_targets, no_obj_targets, class_targets, box_targets = targets\n",
+    "        objectness_loss = self.bce_loss(obj_preds, obj_targets)\n",
+    "\n",
+    "        no_objectness_loss = self.bce_loss(no_obj_preds, no_obj_targets)\n",
+    "\n",
+    "        classification_loss = self.ce_loss(class_preds, class_targets)\n",
+    "\n",
+    "        box_loss = self.mse_loss(box_preds, box_targets)\n",
+    "\n",
+    "        # Combine losses\n",
+    "        total_loss = (\n",
+    "            objectness_loss + \n",
+    "            no_objectness_loss + \n",
+    "            classification_loss + \n",
+    "            box_loss\n",
+    "        )\n",
+    "\n",
+    "        return total_loss"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "2e6c45fb-87bc-430f-be57-fe68aa915242",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Optimizer\n",
+    "optimizer = torch.optim.Adam(model.parameters(), lr=0.001)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "ef53d453-b3fa-430a-826e-39d21e0882c2",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Training Loop\n",
+    "num_epochs = 10  \n",
+    "for epoch in range(num_epochs):\n",
+    "    for images, targets in data_loader:\n",
+    "        optimizer.zero_grad()\n",
+    "        \n",
+    "        outputs = model(images)\n",
+    "        \n",
+    "        loss = loss_function(outputs, targets)\n",
+    "        \n",
+    "        loss.backward()\n",
+    "        optimizer.step()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "2476c91f-a0ea-44b7-9555-2126a1a14666",
+   "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": "3.12.2"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

documents/PoC/Mustafa/data.yaml

@@ -0,0 +1,113 @@
+train: ../train/images
+val: ../test/images
+
+nc: 127
+names:
+  [
+    "Paper Cup",
+    "Snack or Candy Bag or Wrapper ",
+    "Wipe",
+    "Wax Paper",
+    "Latex Gloves",
+    "Juice or Other Pouch",
+    "Diaper",
+    "Padded Envelope (mixed materials)",
+    "Blister Pack",
+    "Pens and Pencils",
+    "Miscellaneous Office Supplies",
+    "Facemask and Other PPE",
+    "Shelf Stable Carton",
+    "Soiled Plastic",
+    "Soiled Metal",
+    "Soiled Glass",
+    "Ceramics",
+    "Unclassifiable",
+    "Filled Bag",
+    "Coffee Pod",
+    "Other Trash",
+    "Flexible container lid / seal",
+    "Snack Food Canister",
+    "Hard Cover Books",
+    "Compostable Fiber Ware",
+    "Compostable Cutlery",
+    "Compostable Plastic Cups",
+    "Compostable Paper Cups",
+    "Paper Towel/Napkins/Tissue",
+    "Wooden Coffee Stirrer or Chopstick",
+    "Soiled Cardboard Box",
+    "Compostable Plastic Lid",
+    "Food Soiled Paper",
+    "Other compostable material",
+    "Sandwich paper wrapper",
+    "Plastic strapping",
+    "Batteries",
+    "Cables",
+    "Computers",
+    "Monitors",
+    "Toner and Ink Cartridges",
+    "Miscellaneous Electronics",
+    "LED Lightbulb",
+    "Meat and Fish",
+    "Bones and Shells",
+    "Cheese and Other Fats",
+    "Fruits And Veggies",
+    "Other Food or Mixed Food",
+    "Breads",
+    "Grains",
+    "Tea Bags",
+    "Coffee Grounds",
+    "Egg Shell",
+    "Glass Bottles",
+    "Glass Jars",
+    "Broken Glass",
+    "Other Clean Glass",
+    "Drinking glass or glass ovenware",
+    "Metal Can",
+    "Aluminum Foil",
+    "Aluminum Catering Tray",
+    "Other Clean Metal",
+    "Aerosol Can",
+    "Metallic Bottle Cap or Lid",
+    "Metal Strapping",
+    "Liquids",
+    "Leaves, Flowers, Grass Clippings",
+    "Office Paper",
+    "Shredded Paper",
+    "Clean Cardboard",
+    "Refrigerated Beverage Carton",
+    "Magazines Newspaper",
+    "Receipts and Thermal Paper",
+    "Empty Paper Bag",
+    "Cardboard Coffee Cup Sleeve",
+    "Clean Paper Plate",
+    "Colored Memo Note",
+    "Office Folder",
+    "Paper Roll",
+    "Wrapping Paper",
+    "Other Clean Paper",
+    "Plastic Drink Bottle",
+    "Plastic Milk Jug or Personal Care Bottle",
+    "Empty Plastic Bag",
+    "Yogurt Tub or Container",
+    "Expanded Polystyrene (styrofoam)",
+    "Other Clean Plastics (rigid)",
+    "Straws",
+    "Clear Clamshell Container",
+    "Plastic Cutlery",
+    "Plastic Lid except black",
+    "Plastic Coffee Stirrer",
+    "Clear Plastic Cup",
+    "Colored Plastic Cup",
+    "Black Plastic",
+    "Plastic Wrap",
+    "Bubble Wrap",
+    "Incandescent Lightbulbs",
+    "CFL Lightbulbs",
+    "Textiles and Clothes",
+    "Food canister",
+    "Plastic Lid",
+    "Other Trash",
+    "Paper Wrapper",
+    "Black Plastic Container",
+    "Padded Envelope",
+  ]