{"metadata":{"kernelspec":{"name":"python3","display_name":"Python 3","language":"python"},"language_info":{"name":"python","version":"3.10.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"none","dataSources":[{"sourceId":67836,"databundleVersionId":7527339,"sourceType":"competition"}],"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":false},"papermill":{"default_parameters":{},"duration":1012.214546,"end_time":"2024-01-20T13:05:55.388071","environment_variables":{},"exception":null,"input_path":"__notebook__.ipynb","output_path":"__notebook__.ipynb","parameters":{},"start_time":"2024-01-20T12:49:03.173525","version":"2.4.0"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# [DL Enigma 1.0 - SUST CSE Carnival 2024](https://www.kaggle.com/competitions/dl-enigma-10-sust-cse-carnival-2024/overview)\n**Bangladesh Road Object Detection for Autonomous Vehicles Challenge**\n\n\n","metadata":{"papermill":{"duration":0.006768,"end_time":"2024-01-20T12:49:06.697217","exception":false,"start_time":"2024-01-20T12:49:06.690449","status":"completed"},"tags":[]}},{"cell_type":"markdown","source":"# πŸ” Key Points\n* Use of **external datasets** to train the model is **prohibited**. Publicly available **pre-training models** are **allowed**, the data used for pre-training the models must be disclose.\n* The dataset comprises 9825 images covering 9 districts in Bangladesh, with 78,943 objects annotated with rectangular bounding boxes across **13 classes**, presenting diverse road types for algorithmic challenges in various driving contexts.\n* The competition focuses on evaluating **mAP50** with IoU set at 0.5, emphasizing the significance of detecting all 13 classes for optimal performance.","metadata":{}},{"cell_type":"markdown","source":"# πŸ—οΈ Install Libraries","metadata":{}},{"cell_type":"code","source":"!pip install ultralytics","metadata":{"_kg_hide-input":false,"_kg_hide-output":true,"execution":{"iopub.status.busy":"2024-02-12T16:05:58.930224Z","iopub.execute_input":"2024-02-12T16:05:58.930596Z","iopub.status.idle":"2024-02-12T16:06:14.549892Z","shell.execute_reply.started":"2024-02-12T16:05:58.930567Z","shell.execute_reply":"2024-02-12T16:06:14.548763Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ“₯ Import Libraries","metadata":{}},{"cell_type":"code","source":"import os\nimport numpy as np\nimport pandas as pd\nfrom tqdm import tqdm\n\nimport random\n\nfrom os import listdir\nfrom os.path import isfile,join\n\nfrom ultralytics import RTDETR\n\n\nimport matplotlib.pyplot as plt\nimport matplotlib.patches as patches\nfrom PIL import Image","metadata":{"papermill":{"duration":0.832795,"end_time":"2024-01-20T12:49:07.54759","exception":false,"start_time":"2024-01-20T12:49:06.714795","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2024-02-12T16:06:14.551889Z","iopub.execute_input":"2024-02-12T16:06:14.552227Z","iopub.status.idle":"2024-02-12T16:06:22.535212Z","shell.execute_reply.started":"2024-02-12T16:06:14.552195Z","shell.execute_reply":"2024-02-12T16:06:22.534042Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ‹ A Dive into MetaData\n* `train/images`- Folder contains training images of format `{District}_{Location/Time}_{image_id}.jpg`\n* `train/labels`- Folder contains training labels of format `{District}_{Location/Time}_{image_id}.txt`. Labels are in **YOLO** format so this `.txt` files contain classes and bounding boxes accordingly.\n* `District`- 9 districts in Bangladesh: Sylhet, Dhaka, Rajshahi, Mymensingh, Maowa, Chittagong, Sirajganj, Sherpur, and Khulna.\n* `Location/Time`- Specific Location or time the photo was taken.\n* `image_id`- ID code for the image.\n* **YOLO Format**- one *.txt file per image. The *.txt file should be formatted with one row per object in class x_center y_center width height format. Box coordinates must be in normalized xywh format (from 0 to 1).","metadata":{}},{"cell_type":"markdown","source":"# πŸ› οΈ Essential Functions\n* `showim_xyxy(path, boxes, classes, scores)`- here,\n * path - absolute path of the image file\n * boxes - list containing bounding boxes, here boxes contains all the rectangular box with x_min,\n y_min, x_max, y_max points\n * classes - list containing class of each box\n * scores - list containing confidence score of each box\n \n **This function plots image with predicted bounding boxes**\n*`get_boxes( boxes )`- here,\n * boxes - bounding boxes with coordinate xywh (x_center, y_center, width, height) format\n \n **This function returns value in xyxy (x_min, y_min, x_max, y_max) format**","metadata":{}},{"cell_type":"code","source":"def showim_xyxy(path,boxes,classes,scores):\n image = plt.imread(path)\n fig, ax = plt.subplots()\n ax.imshow(image)\n i=0\n for i in range(len(boxes)):\n x1, y1, x2, y2 = boxes[i]\n imge = Image.open(path) \n w = imge.width \n h = imge.height\n x1=x1*w\n y1=y1*h\n x2=x2*w\n y2=y2*h\n rect = patches.Rectangle((x1, y1), x2 - x1, y2 - y1, linewidth=2, edgecolor='red', facecolor='none')\n ax.add_patch(rect)\n ax.text(x1, y1 - 5, f\"{classes[i]} {scores[i]:.2f}\", fontsize=8, color='red')\n plt.axis('off') # Turn off axes for a cleaner look\n plt.title('Image with Bounding Boxes and Labels')\n plt.show()\n\n\ndef get_boxes(boxes):\n box_in_img = []\n for i, score in enumerate(boxes):\n single_pred_str = \"\"\n \n x_center , y_center, width,height = boxes[i]\n x1 = float(x_center) - (float(width) / 2)\n y1 = float(y_center) - (float(height) / 2)\n width = float(width)\n height = float(height)\n box_in_img.append([x1,y1,x1+width,y1+height])\n return box_in_img","metadata":{"execution":{"iopub.status.busy":"2024-02-12T16:06:28.417751Z","iopub.execute_input":"2024-02-12T16:06:28.418599Z","iopub.status.idle":"2024-02-12T16:06:28.429439Z","shell.execute_reply.started":"2024-02-12T16:06:28.418564Z","shell.execute_reply":"2024-02-12T16:06:28.428379Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ•΅ Exploratory Data Analysis","metadata":{}},{"cell_type":"code","source":"ipath = '/kaggle/input/dl-enigma-10-sust-cse-carnival-2024/dlenigma1/BadODD/images/train'\nimages = [join(ipath,f) for f in listdir(ipath) if isfile(join(ipath, f))]\nlpath = '/kaggle/input/dl-enigma-10-sust-cse-carnival-2024/dlenigma1/BadODD/labels/train'\nlabels = [join(lpath,f) for f in listdir(lpath) if isfile(join(lpath, f))]\nimages.sort()\nlabels.sort()","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-02-12T16:06:34.359474Z","iopub.execute_input":"2024-02-12T16:06:34.35984Z","iopub.status.idle":"2024-02-12T16:06:52.567115Z","shell.execute_reply.started":"2024-02-12T16:06:34.359812Z","shell.execute_reply":"2024-02-12T16:06:52.566302Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"#randomly printing an images with labels\nrandom.seed(10)\nid = random.randint(0, 5000)\nimg = images[id]\nlab = labels[id]\nboxes = []\nclasses = []\nscores = []\nwith open(lab,'r') as file:\n content = file.read().split('\\n')\nfor i in range(len(content)):\n if not content[i].strip():\n continue\n single_box = [float(num) for num in content[i].split(' ')]\n boxes.append(single_box[1:])\n classes.append(single_box[0])\n scores.append(float(1.0))\nshowim_xyxy(img,get_boxes(boxes),classes,scores) #we have to first convert boxes from xywh form to xyxy","metadata":{"execution":{"iopub.status.busy":"2024-02-12T16:06:52.568812Z","iopub.execute_input":"2024-02-12T16:06:52.569076Z","iopub.status.idle":"2024-02-12T16:06:53.201129Z","shell.execute_reply.started":"2024-02-12T16:06:52.569053Z","shell.execute_reply":"2024-02-12T16:06:53.200086Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"### Class And Bounding Boxes","metadata":{}},{"cell_type":"code","source":"class_id_to_name = {\n 0: 'auto_rickshaw',\n 1: 'bicycle',\n 2: 'bus',\n 3: 'car',\n 4: 'cart_vehicle',\n 5: 'construction_vehicle',\n 6: 'motorbike',\n 7: 'person',\n 8: 'priority_vehicle',\n 9: 'three_wheeler',\n 10: 'train',\n 11: 'truck',\n 12: 'wheelchair'\n}\n\n# Define the label folder path\nlabel_folder = \"/kaggle/input/dl-enigma-10-sust-cse-carnival-2024/dlenigma1/BadODD/labels/train\"\n\n# Initialize a dictionary to store class counts\nclass_counts = {class_id: 0 for class_id in class_id_to_name}\n\n# Iterate through each text file in the label folder\nfor filename in os.listdir(label_folder):\n if filename.endswith(\".txt\"):\n file_path = os.path.join(label_folder, filename)\n \n # Read the content of the label file and count class occurrences\n with open(file_path, 'r') as file:\n lines = file.readlines()\n\n for line in lines:\n parts = line.strip().split()\n if parts: # Check if the line is not empty\n class_id = int(parts[0])\n if class_id in class_counts:\n class_counts[class_id] += 1\n\n# Print the class counts\nfor class_id, count in class_counts.items():\n class_name = class_id_to_name[class_id]\n print(f'Class {class_id} ({class_name}): {count} instances')\nimport numpy as np\n# Create a bar plot\nclass_ids = np.array(list(class_counts.keys()))\n# print(class_ids)\nclass_names = [class_id_to_name[class_id] for class_id in class_ids]\ncounts = np.array(list(class_counts.values()))\ncolors = plt.cm.viridis(np.linspace(0, 1, len(class_ids)))\n\nplt.bar(class_names, counts, color=colors)\nplt.xlabel('Class')\nplt.ylabel('Number of Instances')\nplt.title('Number of Instances in Each Class')\nplt.xticks(rotation=45, ha=\"right\") # Rotate x-axis labels for better visibility\nplt.tight_layout()\nplt.show()","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-02-12T16:06:53.202352Z","iopub.execute_input":"2024-02-12T16:06:53.202638Z","iopub.status.idle":"2024-02-12T16:07:17.687534Z","shell.execute_reply.started":"2024-02-12T16:06:53.202612Z","shell.execute_reply":"2024-02-12T16:07:17.686595Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"### Day/Night Image","metadata":{}},{"cell_type":"code","source":"# pie plot of distribution of day and night images\nnight_images = [filename for filename in images if 'night' in filename.lower()]\nnight_labels = [filename for filename in labels if 'night' in filename.lower()]\nlen(night_images)\nlabel = ['Day', 'Night']\nsizes = [len(images)-len(night_images),len(night_images)]\nplt.pie(sizes, labels=label, autopct='%1.1f%%', startangle=90)\nplt.title('Distribution of Day and Night')\nplt.show()\n\n# class distribution of day and night images\n\nnight_class_counts = {class_id: 0 for class_id in class_id_to_name}\n\nfor file_path in night_labels:\n # Read the content of the label file and count class occurrences\n with open(file_path, 'r') as file:\n lines = file.readlines()\n\n for line in lines:\n parts = line.strip().split()\n if parts: # Check if the line is not empty\n class_id = int(parts[0])\n if class_id in night_class_counts:\n night_class_counts[class_id] += 1\n\nnight_counts = np.array(list(night_class_counts.values()))\nday_counts = counts-night_counts\n\nfig, axes = plt.subplots(nrows=1, ncols=2, figsize=(10, 5))\ncolors = plt.cm.viridis(np.linspace(0, 1, len(class_ids)))\n\naxes[0].bar(class_names, night_counts, color=colors)\naxes[0].set_xlabel('Class')\naxes[0].set_ylabel('Number of Instances')\naxes[0].set_title('Number of Instances in Each Class')\naxes[0].set_xticklabels(class_names, rotation=45, ha=\"right\")\n\naxes[0].set_ylim(0, 17000)\naxes[0].set_ylim(0, 17000)\n\n# Bar plot for day counts\naxes[1].bar(class_names, day_counts, color=colors)\naxes[1].set_xlabel('Class')\naxes[1].set_ylabel('Number of Instances')\naxes[1].set_title('Number of Instances in Each Class (Day)')\naxes[1].set_xticklabels(class_names, rotation=45, ha=\"right\") # Rotate x-axis labels for better visibility\nplt.tight_layout()\nplt.show()","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-02-12T16:07:17.689119Z","iopub.execute_input":"2024-02-12T16:07:17.68942Z","iopub.status.idle":"2024-02-12T16:07:18.968908Z","shell.execute_reply.started":"2024-02-12T16:07:17.689394Z","shell.execute_reply":"2024-02-12T16:07:18.96789Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ“ˆ WandB\nWeights & Biases (WandB) is a machine learning development platform that allows users to track and visualize various aspects of their model training process in real-time. Utilize W&B to accelerate model development by incorporating experiment tracking, dataset versioning, and efficient model management. Some of them are:\n* Monitor, contrast, and depict machine learning experiments.\n* Access real-time metrics, terminal logs, and system statistics on a centralized dashboard as they are continuously streamed.\n* Display improvement graphs for different versions, discuss and address bugs, and showcase progress towards milestones.","metadata":{}},{"cell_type":"code","source":"import wandb\n\ntry:\n from kaggle_secrets import UserSecretsClient\n user_secrets = UserSecretsClient()\n api_key = user_secrets.get_secret(\"WANDB\")\n wandb.login(key=api_key,relogin=True)\n anonymous = None\nexcept:\n wandb.login(anonymous='must')\n print('To use your W&B account,\\nGo to Add-ons -> Secrets and provide your W&B access token. Use the Label name as WANDB. \\nGet your W&B access token from here: https://wandb.ai/authorize')","metadata":{"papermill":{"duration":14.771006,"end_time":"2024-01-20T12:49:22.325184","exception":false,"start_time":"2024-01-20T12:49:07.554178","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2024-02-12T16:12:49.138847Z","iopub.execute_input":"2024-02-12T16:12:49.139297Z","iopub.status.idle":"2024-02-12T16:12:49.364974Z","shell.execute_reply.started":"2024-02-12T16:12:49.139265Z","shell.execute_reply":"2024-02-12T16:12:49.364213Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ“¦Dataset","metadata":{}},{"cell_type":"markdown","source":"## Splitting into Train and Val\nWe will use the full dataset for training as it gives the best score and 20% of total data for validation.","metadata":{}},{"cell_type":"code","source":"import os\nimport shutil\nfrom sklearn.model_selection import train_test_split\nfrom tqdm import tqdm\n\ndef copy_data(images_path, labels_path, destination_path, train_ratio=0.8, random_seed=42):\n # Create new directories for train and val in the destination path\n train_dir = os.path.join(destination_path, 'train')\n val_dir = os.path.join(destination_path, 'val')\n os.makedirs(train_dir, exist_ok=True)\n os.makedirs(val_dir, exist_ok=True)\n\n # Create subdirectories for 'images' and 'labels' within 'train' and 'val'\n train_images_dir = os.path.join(train_dir, 'images')\n train_labels_dir = os.path.join(train_dir, 'labels')\n val_images_dir = os.path.join(val_dir, 'images')\n val_labels_dir = os.path.join(val_dir, 'labels')\n os.makedirs(train_images_dir, exist_ok=True)\n os.makedirs(train_labels_dir, exist_ok=True)\n os.makedirs(val_images_dir, exist_ok=True)\n os.makedirs(val_labels_dir, exist_ok=True)\n\n # List all image and label files\n image_files = os.listdir(os.path.join(images_path, 'train'))\n label_files = os.listdir(os.path.join(labels_path, 'train'))\n\n # Split files into train and val\n image_train, image_val, label_train, label_val = train_test_split(\n image_files, label_files, test_size=(1 - train_ratio), random_state=random_seed\n )\n\n # Copy images and labels to 'images' and 'labels' folders in train\n for image_file in tqdm(image_files, desc='Copying train images'):\n shutil.copy2(os.path.join(images_path, 'train', image_file), os.path.join(train_images_dir, image_file))\n for label_file in tqdm(label_files, desc='Copying train labels'):\n shutil.copy2(os.path.join(labels_path, 'train', label_file), os.path.join(train_labels_dir, label_file))\n\n # Copy remaining images and labels to 'images' and 'labels' folders in val\n for image_file in tqdm(image_val, desc='Copying val images'):\n shutil.copy2(os.path.join(images_path, 'train', image_file), os.path.join(val_images_dir, image_file))\n for label_file in tqdm(label_val, desc='Copying val labels'):\n shutil.copy2(os.path.join(labels_path, 'train', label_file), os.path.join(val_labels_dir, label_file))\n\nif __name__ == \"__main__\":\n images_folder = '/kaggle/input/dl-enigma-10-sust-cse-carnival-2024/dlenigma1/BadODD/images'\n labels_folder = '/kaggle/input/dl-enigma-10-sust-cse-carnival-2024/dlenigma1/BadODD/labels'\n destination_folder = '/kaggle/working/datasets/badodd/'\n\n copy_data(images_folder, labels_folder, destination_folder)","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-02-12T16:10:33.858086Z","iopub.execute_input":"2024-02-12T16:10:33.858465Z","iopub.status.idle":"2024-02-12T16:11:53.540335Z","shell.execute_reply.started":"2024-02-12T16:10:33.858434Z","shell.execute_reply":"2024-02-12T16:11:53.539344Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"import os\nimport matplotlib.pyplot as plt\n\n# Define the paths to the image folders\ntrain_path = \"/kaggle/working/datasets/badodd/train/images\"\nval_path = \"/kaggle/working/datasets/badodd/val/images\"\n\n# Function to count the number of images in a folder\ndef count_images(folder_path):\n return len([f for f in os.listdir(folder_path) if os.path.isfile(os.path.join(folder_path, f))])\n\n# Count the number of images in each folder\ntrain_images_count = count_images(train_path)\nval_images_count = count_images(val_path)\n\n# Bar plot\nfolders = ['Train', 'Validation']\nimage_counts = [train_images_count, val_images_count]\n\nplt.bar(folders, image_counts, color=['blue', 'orange'])\nplt.xlabel('Folders')\nplt.ylabel('Number of Images')\nplt.title('Number of Images in Train and Validation Folders')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2024-02-12T16:11:53.541848Z","iopub.execute_input":"2024-02-12T16:11:53.542403Z","iopub.status.idle":"2024-02-12T16:11:53.767247Z","shell.execute_reply.started":"2024-02-12T16:11:53.542375Z","shell.execute_reply":"2024-02-12T16:11:53.766122Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ”© Configuration","metadata":{}},{"cell_type":"code","source":"file_content = \"\"\"\npath: /kaggle/working/datasets/badodd # dataset root dir\ntrain: train # train images\nval: val # val images\n\n# Classes\nnames:\n 0: auto_rickshaw\n 1: bicycle\n 2: bus\n 3: car\n 4: cart_vehicle\n 5: construction_vehicle\n 6: motorbike\n 7: person\n 8: priority_vehicle\n 9: three_wheeler\n 10: train\n 11: truck\n 12: wheelchair\n\"\"\"\n\nwith open(\"rtdetr.yaml\", mode=\"w\") as f:\n f.write(file_content)","metadata":{"papermill":{"duration":0.016909,"end_time":"2024-01-20T12:49:34.432163","exception":false,"start_time":"2024-01-20T12:49:34.415254","status":"completed"},"tags":[],"execution":{"iopub.status.busy":"2024-02-12T16:11:53.768432Z","iopub.execute_input":"2024-02-12T16:11:53.76874Z","iopub.status.idle":"2024-02-12T16:11:53.77534Z","shell.execute_reply.started":"2024-02-12T16:11:53.768713Z","shell.execute_reply":"2024-02-12T16:11:53.774228Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ‹οΈβ€β™‚οΈTraining\n### Some notes about training\n\n* We use **RTDETR** for this competetion.\n* Train model1 for 100 epochs and model2 for 60 epochs.\n* Then we combine this two models with **WBF**\n* Using **GPU T4x2** as **Accelerator** given by Kaggle for faster training\n\n### RTDETR\n\nReal-Time Detection Transformer (RT-DETR), developed by Baidu, is a cutting-edge end-to-end object detector that provides real-time performance while maintaining high accuracy. It leverages the power of Vision Transformers (ViT) to efficiently process multiscale features by decoupling intra-scale interaction and cross-scale fusion. RT-DETR is highly adaptable, supporting flexible adjustment of inference speed using different decoder layers without retraining. The model excels on accelerated backends like CUDA with TensorRT, outperforming many other real-time object detectors.\n\n\n### Parameters\n * `epochs` : number of epochs\n * `imgsz` : image_size\n * `batch` : batch_size\n * `pretrained` : using pretrained weight (True/False)\n * `data` : config of dataset\n * `seed` : random seed for regeneration\n * `device` : Accelerator id for training\n * `cache` : load all images into the memory, will speed up the training\n * `Augmentations & others` : degrees, translate, scale, etc.(Here we prefered to use default augmentations mostly)","metadata":{}},{"cell_type":"markdown","source":"### model1 with 100 epochs","metadata":{}},{"cell_type":"code","source":"model1 = RTDETR('rtdetr-x.pt')\nmodel1.train(\n \n #loading data and model parameters\n data='/kaggle/working/rtdetr.yaml',\n pretrained=True,\n cache=True, \n device=[0, 1],\n \n #training parameteres\n epochs=100,\n batch=16, \n imgsz=640, \n save=True,\n \n #random seed\n seed = 0,\n \n # Augmentation\n degrees=0.0, \n translate=0.1,\n scale=0.5, \n shear=0.0, \n perspective=0.0,\n flipud=0.0, \n fliplr=0.5, \n mosaic=1.0, \n mixup=0.0, \n\n # Optimization parameters\n lr0=0.018,\n patience=3,\n optimizer=\"auto\",\n momentum=0.947,\n weight_decay=0.0005,\n warmup_momentum=0.8,\n)","metadata":{"papermill":{"duration":1.445596,"end_time":"2024-01-20T12:49:35.884406","exception":false,"start_time":"2024-01-20T12:49:34.43881","status":"completed"},"tags":[],"_kg_hide-output":true,"execution":{"iopub.status.busy":"2024-02-12T16:13:08.803995Z","iopub.execute_input":"2024-02-12T16:13:08.804382Z","iopub.status.idle":"2024-02-12T17:20:52.47572Z","shell.execute_reply.started":"2024-02-12T16:13:08.804348Z","shell.execute_reply":"2024-02-12T17:20:52.474847Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"### model2 with 60 epochs","metadata":{}},{"cell_type":"code","source":"model2 = RTDETR('rtdetr-x.pt')\nmodel2.train(\n \n #loading data and model parameters\n data='/kaggle/working/rtdetr.yaml',\n pretrained=True,\n cache=True, \n device=[0, 1],\n \n #training parameteres\n epochs=60,\n batch=16, \n imgsz=640, \n save=True,\n \n #random seed\n seed = 0,\n \n # Augmentation\n degrees=0.0, \n translate=0.1,\n scale=0.5, \n shear=0.0, \n perspective=0.0,\n flipud=0.0, \n fliplr=0.5, \n mosaic=1.0, \n mixup=0.0, \n\n # Optimization parameters\n lr0=0.018,\n patience=3,\n optimizer=\"auto\",\n momentum=0.947,\n weight_decay=0.0005,\n warmup_momentum=0.8,\n)","metadata":{"execution":{"iopub.execute_input":"2024-01-20T13:05:48.77935Z","iopub.status.busy":"2024-01-20T13:05:48.778957Z","iopub.status.idle":"2024-01-20T13:05:48.785209Z","shell.execute_reply":"2024-01-20T13:05:48.784298Z"},"papermill":{"duration":0.368677,"end_time":"2024-01-20T13:05:48.787504","exception":false,"start_time":"2024-01-20T13:05:48.418827","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ”­ Results","metadata":{}},{"cell_type":"markdown","source":"### Batch Image","metadata":{}},{"cell_type":"code","source":"plt.figure(figsize = (10,10))\nplt.axis('off')\nplt.imshow(plt.imread(f'/kaggle/working/runs/detect/train/labels.jpg'));","metadata":{"execution":{"iopub.status.busy":"2024-02-12T17:20:52.477967Z","iopub.execute_input":"2024-02-12T17:20:52.478752Z","iopub.status.idle":"2024-02-12T17:20:53.184202Z","shell.execute_reply.started":"2024-02-12T17:20:52.478711Z","shell.execute_reply":"2024-02-12T17:20:53.183209Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"### Class Distribution","metadata":{}},{"cell_type":"code","source":"plt.figure(figsize = (10,10))\nplt.axis('off')\nplt.imshow(plt.imread(f'/kaggle/working/runs/detect/train/labels_correlogram.jpg'));","metadata":{"execution":{"iopub.status.busy":"2024-02-12T17:20:53.185454Z","iopub.execute_input":"2024-02-12T17:20:53.185789Z","iopub.status.idle":"2024-02-12T17:20:54.08462Z","shell.execute_reply.started":"2024-02-12T17:20:53.185759Z","shell.execute_reply":"2024-02-12T17:20:54.083735Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"### Score Vs Epoch","metadata":{}},{"cell_type":"code","source":"plt.figure(figsize=(30,15))\nplt.axis('off')\nplt.imshow(plt.imread(f'/kaggle/working/runs/detect/train/results.png'));","metadata":{"execution":{"iopub.status.busy":"2024-02-12T17:20:54.0866Z","iopub.execute_input":"2024-02-12T17:20:54.086908Z","iopub.status.idle":"2024-02-12T17:20:55.572791Z","shell.execute_reply.started":"2024-02-12T17:20:54.08688Z","shell.execute_reply":"2024-02-12T17:20:55.57178Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"### Confusion Matrix","metadata":{}},{"cell_type":"code","source":"plt.figure(figsize=(12,10))\nplt.axis('off')\nplt.imshow(plt.imread(f'/kaggle/working/runs/detect/train/confusion_matrix.png'));","metadata":{"execution":{"iopub.status.busy":"2024-02-12T17:20:55.57411Z","iopub.execute_input":"2024-02-12T17:20:55.574509Z","iopub.status.idle":"2024-02-12T17:20:57.219132Z","shell.execute_reply.started":"2024-02-12T17:20:55.574474Z","shell.execute_reply":"2024-02-12T17:20:57.218234Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"### Metrics","metadata":{}},{"cell_type":"code","source":"for metric in ['F1', 'PR', 'P', 'R']:\n print(f'Metric: {metric}')\n plt.figure(figsize=(12,10))\n plt.axis('off')\n plt.imshow(plt.imread(f'/kaggle/working/runs/detect/train/{metric}_curve.png'));\n plt.show()","metadata":{"execution":{"iopub.status.busy":"2024-02-12T17:20:57.220574Z","iopub.execute_input":"2024-02-12T17:20:57.220923Z","iopub.status.idle":"2024-02-12T17:20:59.874278Z","shell.execute_reply.started":"2024-02-12T17:20:57.220894Z","shell.execute_reply":"2024-02-12T17:20:59.873288Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# πŸ“’ References and Acknowledgement\n* Some ideas of this notebook is form [Great-Barrier-Reef: YOLOv5 [train] 🌊](https://www.kaggle.com/awsaf49/great-barrier-reef-yolov5-train)\n* Image 1 is taken from the [Competition](https://www.kaggle.com/competitions/dl-enigma-10-sust-cse-carnival-2024/overview) main page\n* Image 2 is taken from ultralytics [RTDETR Docs](https://docs.ultralytics.com/models/rtdetr/)\n* Image 3 is from [This blog](https://m.somewhereinblog.net/)\n\n
\n \"Your\n
","metadata":{}}]}