lacmus_detection.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Lacmus Demo",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/ipovalyaev/events/blob/collab_detection_notebook/lacmus_detection.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "t6MPjfT5NrKQ"
      },
      "source": [
        "This notebook is intended to demonstrate and test **Lacmus** neural network, based on Yolov5 and trained on Lacmus dataset. \n",
        "More information available at  https://lacmus-foundation.github.io  \n",
        "\n",
        "!!! \n",
        "\n",
        "This collab is not intended to use in S&R operation - it's not suited for fast photos processing at large quantities. It will let you test how **Lacmus** network will perform on the photos you have, so you can decide if it works well for your case.\n",
        "\n",
        "For S&R operations you can use our application https://github.com/lacmus-foundation/lacmus or server https://github.com/lacmus-foundation/lacmus-web-api\n",
        "\n",
        "!!!\n",
        "\n",
        "\n",
        "This notebook is based on **YOLOv5 🚀 notebook** by **Ultralytics**, which is freely available for redistribution under the [GPL-3.0 license](https://choosealicense.com/licenses/gpl-3.0/). *italicized text*\n",
        "For more information please visit https://github.com/ultralytics/yolov5 and https://ultralytics.com. Thank you!"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7mGmQbAO5pQb"
      },
      "source": [
        "# Setup\n",
        "\n",
        "Clone repo, install dependencies and check PyTorch and GPU."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wbvMlHd_QwMG"
      },
      "source": [
        "!git clone https://github.com/ultralytics/yolov5  # clone\n",
        "%cd yolov5\n",
        "%pip install -qr requirements.txt  # install\n",
        "\n",
        "import torch\n",
        "from yolov5 import utils\n",
        "display = utils.notebook_init()  # checks"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# We have to download Lacmus model weights. \n",
        "!wget https://lacmus-dvc.fra1.digitaloceanspaces.com/public/weights/yolo5_fullDS_native.pt"
      ],
      "metadata": {
        "id": "VnAVHD-qQS3o"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4JnkELT0cIJg"
      },
      "source": [
        "# 1. Downloading images and running the network\n",
        "\n",
        "`detect.py` runs YOLOv5 inference on a variety of sources and saving results to `runs/detect`. "
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Remove the original Yolo5 examples there, as nature of drone footage is completely different\n",
        "!rm data/images/bus.jpg\n",
        "!rm data/images/zidane.jpg"
      ],
      "metadata": {
        "id": "S8XgvtJlGhQm"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# We use examples photos from our dataset. \n",
        "# You can try, before running a network download those by links below and try to find people by yourself ;)\n",
        "!wget https://lacmus-dvc.fra1.digitaloceanspaces.com/public/examples/418.jpg\n",
        "!wget https://lacmus-dvc.fra1.digitaloceanspaces.com/public/examples/417.jpg\n",
        "!cp 41*.jpg data/images"
      ],
      "metadata": {
        "id": "r9-E_Lp7FzQG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "You can upload your photo by uncommenting content of 2 cells below and adjusting the filename to the one was uploaded. \n",
        "Photo should be taken from approx 50 meters height and have approx 4000x3000 resolution."
      ],
      "metadata": {
        "id": "uoeKWMzAVkq-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# from google.colab import files\n",
        "# uploaded = files.upload()"
      ],
      "metadata": {
        "id": "526zVnvSUmXZ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# !cp <PUT_NAME_OF_FILE_YOU_UPLODADED_THERE> data/images"
      ],
      "metadata": {
        "id": "AG95U_oFMScF"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# If you run detection several times Yolo creates folders like runs/detect/exp2, runs/detect/exp3, etc. \n",
        "# To avoid this remove the exp folder before running detection\n",
        "!rm -rf runs/detect/exp"
      ],
      "metadata": {
        "id": "oPMVCQfyLkpg"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zR9ZbuQCH7FX"
      },
      "source": [
        "!python detect.py --weights yolo5_fullDS_native.pt --imgsz 1984 --conf-thres 0.05 --iou-thres 0.01 --source data/images\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "All detection results are in runs/detect/exp/ , including your file, if you've uploaded any, so you can display it as shown below.\n",
        "Please note, that confidence threshold intentionally set low, so this can cause some false positive, but reduces probability of missing a human.\n",
        "Yet you can adjust parameter  `conf-thres` in the cell above and re-running to above cells"
      ],
      "metadata": {
        "id": "cwjXtZpKVV1A"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "display.Image(filename='runs/detect/exp/417.jpg', width=2000)"
      ],
      "metadata": {
        "id": "odFQuRk3Gx_V"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "display.Image(filename='runs/detect/exp/418.jpg', width=2000)"
      ],
      "metadata": {
        "id": "hJvrri0LJt2M"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "In case you found network doesn't detects humans on your photos, please consider submitting your photos to improve our training dataset by uploading photo to any public cloud storage along with verbal instructions how to locate people there or labels in pascal voc format and creating and issue in https://github.com/lacmus-foundation/lacmus-research/issues.\n",
        "\n",
        "For collaboration in R&D or in case you have any questions - please drop a mail to addressed in  https://lacmus-foundation.github.io/#contact"
      ],
      "metadata": {
        "id": "GaPlV92hXACz"
      }
    }
  ]
}