basic-test-yolov5.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e2e94421",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:38.164947Z",
     "iopub.status.busy": "2022-03-22T09:14:38.162631Z",
     "iopub.status.idle": "2022-03-22T09:14:38.185840Z",
     "shell.execute_reply": "2022-03-22T09:14:38.186411Z",
     "shell.execute_reply.started": "2022-03-22T09:10:23.571704Z"
    },
    "papermill": {
     "duration": 0.085337,
     "end_time": "2022-03-22T09:14:38.186859",
     "exception": false,
     "start_time": "2022-03-22T09:14:38.101522",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import os\n",
    "os.listdir('/kaggle/input/tensorflow-great-barrier-reef/train_images/')\n",
    "# modeldir = '../input/frank2022-02-15-ep20'\n",
    "modeldir = '../input/frank20220228b16e100s'\n",
    "model = modeldir + \"/best.pt\"\n",
    "conf_thres=0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "14f1cef7",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:38.265172Z",
     "iopub.status.busy": "2022-03-22T09:14:38.264213Z",
     "iopub.status.idle": "2022-03-22T09:14:38.280144Z",
     "shell.execute_reply": "2022-03-22T09:14:38.281186Z",
     "shell.execute_reply.started": "2022-03-22T09:10:35.080583Z"
    },
    "papermill": {
     "duration": 0.057133,
     "end_time": "2022-03-22T09:14:38.281462",
     "exception": false,
     "start_time": "2022-03-22T09:14:38.224329",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best.pt']"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "os.listdir(modeldir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "fd68a64a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:38.358805Z",
     "iopub.status.busy": "2022-03-22T09:14:38.357839Z",
     "iopub.status.idle": "2022-03-22T09:14:39.235076Z",
     "shell.execute_reply": "2022-03-22T09:14:39.234409Z",
     "shell.execute_reply.started": "2022-03-22T09:10:37.074094Z"
    },
    "papermill": {
     "duration": 0.918259,
     "end_time": "2022-03-22T09:14:39.235249",
     "exception": false,
     "start_time": "2022-03-22T09:14:38.316990",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Arial.ttf.bk        __output__.json   __resultx__.html  \u001b[0m\u001b[01;34myolov5\u001b[0m/\r\n",
      "__notebook__.ipynb  __results__.html  custom.css\r\n"
     ]
    }
   ],
   "source": [
    "%ls ../input/install-yolov5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "4cb52b2b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:39.286159Z",
     "iopub.status.busy": "2022-03-22T09:14:39.285138Z",
     "iopub.status.idle": "2022-03-22T09:14:40.033102Z",
     "shell.execute_reply": "2022-03-22T09:14:40.032456Z",
     "shell.execute_reply.started": "2022-03-22T09:10:39.568551Z"
    },
    "papermill": {
     "duration": 0.776257,
     "end_time": "2022-03-22T09:14:40.033260",
     "exception": false,
     "start_time": "2022-03-22T09:14:39.257003",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "!ln  -s ../input/install-yolov5/yolov5 yolov5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "393850aa",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:40.084266Z",
     "iopub.status.busy": "2022-03-22T09:14:40.081680Z",
     "iopub.status.idle": "2022-03-22T09:14:40.852338Z",
     "shell.execute_reply": "2022-03-22T09:14:40.851607Z",
     "shell.execute_reply.started": "2022-03-22T09:10:42.149520Z"
    },
    "papermill": {
     "duration": 0.795626,
     "end_time": "2022-03-22T09:14:40.852525",
     "exception": false,
     "start_time": "2022-03-22T09:14:40.056899",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "%mkdir  /root/.config/Ultralytics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "4ccf22f9",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:40.899757Z",
     "iopub.status.busy": "2022-03-22T09:14:40.898644Z",
     "iopub.status.idle": "2022-03-22T09:14:41.666892Z",
     "shell.execute_reply": "2022-03-22T09:14:41.666254Z",
     "shell.execute_reply.started": "2022-03-22T09:10:44.858109Z"
    },
    "papermill": {
     "duration": 0.792573,
     "end_time": "2022-03-22T09:14:41.667056",
     "exception": false,
     "start_time": "2022-03-22T09:14:40.874483",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "!cp ../input/install-yolov5/Arial.ttf.bk /root/.config/Ultralytics/Arial.ttf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "61093326",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:41.739476Z",
     "iopub.status.busy": "2022-03-22T09:14:41.738551Z",
     "iopub.status.idle": "2022-03-22T09:14:45.372514Z",
     "shell.execute_reply": "2022-03-22T09:14:45.371545Z",
     "shell.execute_reply.started": "2022-03-22T09:12:43.429719Z"
    },
    "papermill": {
     "duration": 3.683746,
     "end_time": "2022-03-22T09:14:45.372704",
     "exception": false,
     "start_time": "2022-03-22T09:14:41.688958",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import torch\n",
    "import numpy as np\n",
    "\n",
    "from skimage import io  # To Read image\n",
    "from yolov5.detect import *\n",
    "from models.common import DetectMultiBackend\n",
    "from utils.augmentations import letterbox\n",
    "from utils.datasets import LoadImages\n",
    "from utils.general import (check_img_size, non_max_suppression, xyxy2xywh)\n",
    "from utils.torch_utils import select_device\n",
    "from skimage import img_as_ubyte\n",
    "\n",
    "class YoloPredictor:\n",
    "  # Convert a bbox record to a bbox line.\n",
    "  def toBBLine(self, r):\n",
    "    return '%f %d %d %d %d' % (r['conf'], r['x'], r['y'], r['width'], r['height'])\n",
    "\n",
    "# Convert the yolo record of {cx, cy, width, height, conf} to the bbox record\n",
    "# of {x, y, width, height, conf}.\n",
    "# Notice in the yolo record, the cx, cy are the center of the bounding box in\n",
    "# the unit of ratio to the width and height of the img_width and img_height.\n",
    "# The unit of the width and height are also in the unit of the ratio to img_width\n",
    "# and img_height.\n",
    "# In the bbbox record, the x,y are the minimum value of x,y and x,y,width,height\n",
    "# are all in unit of pixel.\n",
    "  def yoloToBB(self, r, img_width, img_height):\n",
    "    x = (r['cx'] - r['width']/2.0) * img_width\n",
    "    y = (r['cy'] - r['height']/2.0) * img_height\n",
    "    width = r['width'] * img_width\n",
    "    height = r['height'] * img_height\n",
    "    return {'x': round(x), 'y': round(y),\n",
    "          'width': round(width), 'height': round(height),\n",
    "          'conf': r['conf']}\n",
    "\n",
    "# Read a line from the yolov5 output .txt file and turn into\n",
    "# record o\n",
    "  def parseLine(self, line):\n",
    "    f = line.split(\" \")\n",
    "    return {'cx': float(f[1]), 'cy': float(f[2]),\n",
    "          'width': float(f[3]), 'height': float(f[4]),\n",
    "          'conf': float(f[5])}\n",
    "\n",
    "  def yoloResultToBBox(self, img_width, img_height, pred):\n",
    "    ret = []\n",
    "    for line in pred:\n",
    "      yolo = self.parseLine(line.strip())\n",
    "      bb = self.yoloToBB(yolo, img_width, img_height)\n",
    "      ret.append(self.toBBLine(bb))\n",
    "    single = ' '.join(ret)\n",
    "    return single\n",
    "\n",
    "\n",
    "  def run(self,\n",
    "          img,\n",
    "          imgsz=(640, 640),  # inference size (height, width)\n",
    "          conf_thres=0.25,  # confidence threshold\n",
    "          iou_thres=0.45,  # NMS IOU threshold\n",
    "          ):\n",
    "\n",
    "    # Load model\n",
    "    stride, names, pt = self.model_.stride, self.model_.names, self.model_.pt\n",
    "    imgsz = check_img_size(imgsz, s=stride)  # check image size\n",
    "\n",
    "    # Run inference\n",
    "    self.model_.warmup(imgsz=(1, 3, *imgsz))  # warmup\n",
    "    imgpath = '/tmp/tt.jpg'\n",
    "\n",
    "    # Somehow we have problem to convert skimage img to cv2 image in memory.\n",
    "    # The jpg compression (qualit=75 as default) is somehow needed.\n",
    "    io.imsave(imgpath, img)\n",
    "    im = cv2.imread(imgpath)\n",
    "\n",
    "    im0s=im\n",
    "    im = letterbox(im, imgsz, stride=stride, auto=True)[0]\n",
    "    im = im.transpose((2, 0, 1))[::-1]  # HWC to CHW, BGR to RGB\n",
    "    im = np.ascontiguousarray(im)\n",
    "    im = torch.from_numpy(im).to(self.device_)\n",
    "    im = im.float()  # uint8 to fp16/32\n",
    "    im /= 255  # 0 - 255 to 0.0 - 1.0\n",
    "    if len(im.shape) == 3:\n",
    "        im = im[None]  # expand for batch dim\n",
    "    # Inference\n",
    "    pred = self.model_(im, augment=False, visualize=False)\n",
    "    # NMS\n",
    "    pred = non_max_suppression(pred, conf_thres, iou_thres, None, False, max_det=1000)\n",
    "    gn = torch.tensor(im0s.shape)[[1, 0, 1, 0]]  # normalization gain whwh\n",
    "    ret = []\n",
    "    #for i, det in enumerate(pred):  # per image\n",
    "    det = pred[0]\n",
    "    if len(det):\n",
    "      # Rescale boxes from img_size to im0 size\n",
    "      det[:, :4] = scale_coords(im.shape[2:], det[:, :4], im0s.shape).round()\n",
    "      for *xyxy, conf, cls in reversed(det):\n",
    "        xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh\n",
    "        line = (cls, *xywh, conf)\n",
    "        ret.append(('%g ' * len(line)).rstrip() % line + '\\n')\n",
    "    return ret\n",
    "\n",
    "  # Predict the item in the img according to img\n",
    "  # Return the string in SPACE separated format\n",
    "  def predict(self, img):\n",
    "    img_height = img.shape[0]\n",
    "    img_width = img.shape[1]\n",
    "\n",
    "    pred = self.run(img=img, imgsz=(img_width, img_height), conf_thres=self.conf_thres_)\n",
    "\n",
    "    return self.yoloResultToBBox(img_width, img_height, pred)\n",
    "\n",
    "  def __init__(self, weights, work_directory, conf_thres = 0.5):\n",
    "    self.work_directory_ = work_directory\n",
    "    self.conf_thres_ = conf_thres\n",
    "    self.device_ = select_device('')\n",
    "    self.model_ = DetectMultiBackend(weights, device=self.device_, dnn=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "15b61dd1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T06:54:40.989932Z",
     "iopub.status.busy": "2022-03-22T06:54:40.989410Z",
     "iopub.status.idle": "2022-03-22T06:54:41.005226Z",
     "shell.execute_reply": "2022-03-22T06:54:41.004246Z",
     "shell.execute_reply.started": "2022-03-22T06:54:40.989873Z"
    },
    "papermill": {
     "duration": 0.022202,
     "end_time": "2022-03-22T09:14:45.416462",
     "exception": false,
     "start_time": "2022-03-22T09:14:45.394260",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "2ea7136c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:45.462058Z",
     "iopub.status.busy": "2022-03-22T09:14:45.460980Z",
     "iopub.status.idle": "2022-03-22T09:14:46.282194Z",
     "shell.execute_reply": "2022-03-22T09:14:46.281064Z",
     "shell.execute_reply.started": "2022-03-22T09:11:09.464861Z"
    },
    "papermill": {
     "duration": 0.845186,
     "end_time": "2022-03-22T09:14:46.282462",
     "exception": false,
     "start_time": "2022-03-22T09:14:45.437276",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "../input/tensorflow-great-barrier-reef/train_images/video_1/1.jpg\r\n"
     ]
    }
   ],
   "source": [
    "%ls ../input/tensorflow-great-barrier-reef/train_images/video_1/1.jpg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "2d000e70",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:46.340006Z",
     "iopub.status.busy": "2022-03-22T09:14:46.339120Z",
     "iopub.status.idle": "2022-03-22T09:14:46.342553Z",
     "shell.execute_reply": "2022-03-22T09:14:46.343583Z",
     "shell.execute_reply.started": "2022-03-22T09:11:12.548123Z"
    },
    "papermill": {
     "duration": 0.035282,
     "end_time": "2022-03-22T09:14:46.343775",
     "exception": false,
     "start_time": "2022-03-22T09:14:46.308493",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from skimage import io  # To Read image\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "9462b932",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:46.400796Z",
     "iopub.status.busy": "2022-03-22T09:14:46.399592Z",
     "iopub.status.idle": "2022-03-22T09:14:47.171273Z",
     "shell.execute_reply": "2022-03-22T09:14:47.170672Z",
     "shell.execute_reply.started": "2022-03-22T09:11:15.437818Z"
    },
    "papermill": {
     "duration": 0.803528,
     "end_time": "2022-03-22T09:14:47.171445",
     "exception": false,
     "start_time": "2022-03-22T09:14:46.367917",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "../input/tensorflow-great-barrier-reef/train_images/video_1/100.jpg\r\n"
     ]
    }
   ],
   "source": [
    "%ls ../input/tensorflow-great-barrier-reef/train_images/video_1/100.jpg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "54f88174",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:47.220727Z",
     "iopub.status.busy": "2022-03-22T09:14:47.219899Z",
     "iopub.status.idle": "2022-03-22T09:14:47.223792Z",
     "shell.execute_reply": "2022-03-22T09:14:47.223191Z",
     "shell.execute_reply.started": "2022-02-28T07:30:29.757377Z"
    },
    "papermill": {
     "duration": 0.030144,
     "end_time": "2022-03-22T09:14:47.223936",
     "exception": false,
     "start_time": "2022-03-22T09:14:47.193792",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# img = io.imread(\"../input/tensorflow-great-barrier-reef/train_images/video_1/100.jpg\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "3d292498",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:47.272390Z",
     "iopub.status.busy": "2022-03-22T09:14:47.271460Z",
     "iopub.status.idle": "2022-03-22T09:14:47.274305Z",
     "shell.execute_reply": "2022-03-22T09:14:47.274985Z",
     "shell.execute_reply.started": "2022-03-22T09:11:21.353371Z"
    },
    "papermill": {
     "duration": 0.029059,
     "end_time": "2022-03-22T09:14:47.275140",
     "exception": false,
     "start_time": "2022-03-22T09:14:47.246081",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "work_directory=\".\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "527dc591",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:47.323002Z",
     "iopub.status.busy": "2022-03-22T09:14:47.322134Z",
     "iopub.status.idle": "2022-03-22T09:14:47.326470Z",
     "shell.execute_reply": "2022-03-22T09:14:47.325941Z",
     "shell.execute_reply.started": "2022-02-28T07:32:19.878172Z"
    },
    "papermill": {
     "duration": 0.030356,
     "end_time": "2022-03-22T09:14:47.326635",
     "exception": false,
     "start_time": "2022-03-22T09:14:47.296279",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# prediction = yoloPredictImage(model, img, work_directory)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "86b46ca8",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:47.373065Z",
     "iopub.status.busy": "2022-03-22T09:14:47.372024Z",
     "iopub.status.idle": "2022-03-22T09:14:47.401300Z",
     "shell.execute_reply": "2022-03-22T09:14:47.400684Z",
     "shell.execute_reply.started": "2022-03-22T09:13:48.311541Z"
    },
    "papermill": {
     "duration": 0.05369,
     "end_time": "2022-03-22T09:14:47.401449",
     "exception": false,
     "start_time": "2022-03-22T09:14:47.347759",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import greatbarrierreef\n",
    "env = greatbarrierreef.make_env()   # initialize the environment\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "525b021d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:47.456788Z",
     "iopub.status.busy": "2022-03-22T09:14:47.455460Z",
     "iopub.status.idle": "2022-03-22T09:14:52.628321Z",
     "shell.execute_reply": "2022-03-22T09:14:52.627701Z",
     "shell.execute_reply.started": "2022-03-22T09:12:50.166579Z"
    },
    "papermill": {
     "duration": 5.20503,
     "end_time": "2022-03-22T09:14:52.628491",
     "exception": false,
     "start_time": "2022-03-22T09:14:47.423461",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "YOLOv5 🚀 v6.0-274-gc43f135 torch 1.9.1 CUDA:0 (Tesla P100-PCIE-16GB, 16281MiB)\n",
      "\n",
      "Fusing layers... \n",
      "Model Summary: 213 layers, 7012822 parameters, 0 gradients\n"
     ]
    }
   ],
   "source": [
    "mymodel = YoloPredictor(model, work_directory)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "8b9a67f2",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-03-22T09:14:52.682800Z",
     "iopub.status.busy": "2022-03-22T09:14:52.681966Z",
     "iopub.status.idle": "2022-03-22T09:14:58.564659Z",
     "shell.execute_reply": "2022-03-22T09:14:58.565421Z",
     "shell.execute_reply.started": "2022-03-22T09:13:53.158943Z"
    },
    "papermill": {
     "duration": 5.913737,
     "end_time": "2022-03-22T09:14:58.565699",
     "exception": false,
     "start_time": "2022-03-22T09:14:52.651962",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "This version of the API is not optimized and should not be used to estimate the runtime of your code on the hidden test set.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING: --img-size (1280, 720) must be multiple of max stride 32, updating to [1280, 736]\n",
      "WARNING: --img-size (1280, 720) must be multiple of max stride 32, updating to [1280, 736]\n",
      "WARNING: --img-size (1280, 720) must be multiple of max stride 32, updating to [1280, 736]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "DONE\n"
     ]
    }
   ],
   "source": [
    "iter_test = env.iter_test()    # an iterator which loops over the test set and sample submission\n",
    "for (pixel_array, sample_prediction_df) in iter_test:\n",
    "    image = pixel_array # [:, :, ::-1]   \n",
    "    prediction = mymodel.predict(image)\n",
    "    sample_prediction_df['annotations'] = prediction\n",
    "    env.predict(sample_prediction_df)   # register your predictions\n",
    "    \n",
    "print(\"DONE\")"
   ]
  }
 ],
 "metadata": {
  "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": "3.7.12"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 32.688719,
   "end_time": "2022-03-22T09:15:00.962526",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2022-03-22T09:14:28.273807",
   "version": "2.3.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}