RGB & TinyYolo with spatial data

This example shows how to run Yolo on the RGB input frame, and how to display both the RGB preview, detections, depth map and spatial information (X,Y,Z). It’s similar to example RGB & MobilenetSSD with spatial data except it is running TinyYolo network. X,Y,Z coordinates are relative to the center of depth map.

setNumClasses - number of YOLO classes setCoordinateSize - size of coordinate setAnchors - yolo anchors setAnchorMasks - anchorMasks26, anchorMasks13 (anchorMasks52 - additionally for full YOLOv4) setIouThreshold - intersection over union threshold setConfidenceThreshold - confidence threshold above which objects are detected

Similar samples:

• Spatial location calculator

• Spatial object tracker on RGB

• RGB & MobilenetSSD with spatial data

• Mono & MobilenetSSD with spatial data

Demo

Setup

Please run the install script to download all required dependencies. Please note that this script must be ran from git context, so you have to download the depthai-python repository first and then run the script

git clone https://github.com/luxonis/depthai-python.git
cd depthai-python/examples
python3 install_requirements.py

For additional information, please follow installation guide

This example script requires external file(s) to run. If you are using:

• depthai-python, run python3 examples/install_requirements.py to download required file(s)

• dephtai-core, required file(s) will get downloaded automatically when building the example

Source code

Python

C++

Also available on GitHub

#!/usr/bin/env python3

from pathlib import Path
import sys
import cv2
import depthai as dai
import numpy as np
import time

'''
Spatial Tiny-yolo example
  Performs inference on RGB camera and retrieves spatial location coordinates: x,y,z relative to the center of depth map.
  Can be used for tiny-yolo-v3 or tiny-yolo-v4 networks
'''

# Get argument first
nnBlobPath = str((Path(__file__).parent / Path('../models/yolo-v4-tiny-tf_openvino_2021.4_6shave.blob')).resolve().absolute())
if 1 < len(sys.argv):
    arg = sys.argv[1]
    if arg == "yolo3":
        nnBlobPath = str((Path(__file__).parent / Path('../models/yolo-v3-tiny-tf_openvino_2021.4_6shave.blob')).resolve().absolute())
    elif arg == "yolo4":
        nnBlobPath = str((Path(__file__).parent / Path('../models/yolo-v4-tiny-tf_openvino_2021.4_6shave.blob')).resolve().absolute())
    else:
        nnBlobPath = arg
else:
    print("Using Tiny YoloV4 model. If you wish to use Tiny YOLOv3, call 'tiny_yolo.py yolo3'")

if not Path(nnBlobPath).exists():
    import sys
    raise FileNotFoundError(f'Required file/s not found, please run "{sys.executable} install_requirements.py"')

# Tiny yolo v3/4 label texts
labelMap = [
    "person",         "bicycle",    "car",           "motorbike",     "aeroplane",   "bus",           "train",
    "truck",          "boat",       "traffic light", "fire hydrant",  "stop sign",   "parking meter", "bench",
    "bird",           "cat",        "dog",           "horse",         "sheep",       "cow",           "elephant",
    "bear",           "zebra",      "giraffe",       "backpack",      "umbrella",    "handbag",       "tie",
    "suitcase",       "frisbee",    "skis",          "snowboard",     "sports ball", "kite",          "baseball bat",
    "baseball glove", "skateboard", "surfboard",     "tennis racket", "bottle",      "wine glass",    "cup",
    "fork",           "knife",      "spoon",         "bowl",          "banana",      "apple",         "sandwich",
    "orange",         "broccoli",   "carrot",        "hot dog",       "pizza",       "donut",         "cake",
    "chair",          "sofa",       "pottedplant",   "bed",           "diningtable", "toilet",        "tvmonitor",
    "laptop",         "mouse",      "remote",        "keyboard",      "cell phone",  "microwave",     "oven",
    "toaster",        "sink",       "refrigerator",  "book",          "clock",       "vase",          "scissors",
    "teddy bear",     "hair drier", "toothbrush"
]

syncNN = True

# Create pipeline
pipeline = dai.Pipeline()

# Define sources and outputs
camRgb = pipeline.create(dai.node.ColorCamera)
spatialDetectionNetwork = pipeline.create(dai.node.YoloSpatialDetectionNetwork)
monoLeft = pipeline.create(dai.node.MonoCamera)
monoRight = pipeline.create(dai.node.MonoCamera)
stereo = pipeline.create(dai.node.StereoDepth)
nnNetworkOut = pipeline.create(dai.node.XLinkOut)

xoutRgb = pipeline.create(dai.node.XLinkOut)
xoutNN = pipeline.create(dai.node.XLinkOut)
xoutDepth = pipeline.create(dai.node.XLinkOut)

xoutRgb.setStreamName("rgb")
xoutNN.setStreamName("detections")
xoutDepth.setStreamName("depth")
nnNetworkOut.setStreamName("nnNetwork")

# Properties
camRgb.setPreviewSize(416, 416)
camRgb.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P)
camRgb.setInterleaved(False)
camRgb.setColorOrder(dai.ColorCameraProperties.ColorOrder.BGR)

monoLeft.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
monoLeft.setCamera("left")
monoRight.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
monoRight.setCamera("right")

# setting node configs
stereo.setDefaultProfilePreset(dai.node.StereoDepth.PresetMode.HIGH_DENSITY)
# Align depth map to the perspective of RGB camera, on which inference is done
stereo.setDepthAlign(dai.CameraBoardSocket.CAM_A)
stereo.setOutputSize(monoLeft.getResolutionWidth(), monoLeft.getResolutionHeight())
stereo.setSubpixel(True)

spatialDetectionNetwork.setBlobPath(nnBlobPath)
spatialDetectionNetwork.setConfidenceThreshold(0.5)
spatialDetectionNetwork.input.setBlocking(False)
spatialDetectionNetwork.setBoundingBoxScaleFactor(0.5)
spatialDetectionNetwork.setDepthLowerThreshold(100)
spatialDetectionNetwork.setDepthUpperThreshold(5000)

# Yolo specific parameters
spatialDetectionNetwork.setNumClasses(80)
spatialDetectionNetwork.setCoordinateSize(4)
spatialDetectionNetwork.setAnchors([10,14, 23,27, 37,58, 81,82, 135,169, 344,319])
spatialDetectionNetwork.setAnchorMasks({ "side26": [1,2,3], "side13": [3,4,5] })
spatialDetectionNetwork.setIouThreshold(0.5)

# Linking
monoLeft.out.link(stereo.left)
monoRight.out.link(stereo.right)

camRgb.preview.link(spatialDetectionNetwork.input)
if syncNN:
    spatialDetectionNetwork.passthrough.link(xoutRgb.input)
else:
    camRgb.preview.link(xoutRgb.input)

spatialDetectionNetwork.out.link(xoutNN.input)

stereo.depth.link(spatialDetectionNetwork.inputDepth)
spatialDetectionNetwork.passthroughDepth.link(xoutDepth.input)
spatialDetectionNetwork.outNetwork.link(nnNetworkOut.input)

# Connect to device and start pipeline
with dai.Device(pipeline) as device:

    # Output queues will be used to get the rgb frames and nn data from the outputs defined above
    previewQueue = device.getOutputQueue(name="rgb", maxSize=4, blocking=False)
    detectionNNQueue = device.getOutputQueue(name="detections", maxSize=4, blocking=False)
    depthQueue = device.getOutputQueue(name="depth", maxSize=4, blocking=False)
    networkQueue = device.getOutputQueue(name="nnNetwork", maxSize=4, blocking=False)

    startTime = time.monotonic()
    counter = 0
    fps = 0
    color = (255, 255, 255)
    printOutputLayersOnce = True

    while True:
        inPreview = previewQueue.get()
        inDet = detectionNNQueue.get()
        depth = depthQueue.get()
        inNN = networkQueue.get()

        if printOutputLayersOnce:
            toPrint = 'Output layer names:'
            for ten in inNN.getAllLayerNames():
                toPrint = f'{toPrint} {ten},'
            print(toPrint)
            printOutputLayersOnce = False

        frame = inPreview.getCvFrame()
        depthFrame = depth.getFrame() # depthFrame values are in millimeters

        depth_downscaled = depthFrame[::4]
        if np.all(depth_downscaled == 0):
            min_depth = 0  # Set a default minimum depth value when all elements are zero
        else:
            min_depth = np.percentile(depth_downscaled[depth_downscaled != 0], 1)
        max_depth = np.percentile(depth_downscaled, 99)
        depthFrameColor = np.interp(depthFrame, (min_depth, max_depth), (0, 255)).astype(np.uint8)
        depthFrameColor = cv2.applyColorMap(depthFrameColor, cv2.COLORMAP_HOT)

        counter+=1
        current_time = time.monotonic()
        if (current_time - startTime) > 1 :
            fps = counter / (current_time - startTime)
            counter = 0
            startTime = current_time

        detections = inDet.detections

        # If the frame is available, draw bounding boxes on it and show the frame
        height = frame.shape[0]
        width  = frame.shape[1]
        for detection in detections:
            roiData = detection.boundingBoxMapping
            roi = roiData.roi
            roi = roi.denormalize(depthFrameColor.shape[1], depthFrameColor.shape[0])
            topLeft = roi.topLeft()
            bottomRight = roi.bottomRight()
            xmin = int(topLeft.x)
            ymin = int(topLeft.y)
            xmax = int(bottomRight.x)
            ymax = int(bottomRight.y)
            cv2.rectangle(depthFrameColor, (xmin, ymin), (xmax, ymax), color, 1)

            # Denormalize bounding box
            x1 = int(detection.xmin * width)
            x2 = int(detection.xmax * width)
            y1 = int(detection.ymin * height)
            y2 = int(detection.ymax * height)
            try:
                label = labelMap[detection.label]
            except:
                label = detection.label
            cv2.putText(frame, str(label), (x1 + 10, y1 + 20), cv2.FONT_HERSHEY_TRIPLEX, 0.5, 255)
            cv2.putText(frame, "{:.2f}".format(detection.confidence*100), (x1 + 10, y1 + 35), cv2.FONT_HERSHEY_TRIPLEX, 0.5, 255)
            cv2.putText(frame, f"X: {int(detection.spatialCoordinates.x)} mm", (x1 + 10, y1 + 50), cv2.FONT_HERSHEY_TRIPLEX, 0.5, 255)
            cv2.putText(frame, f"Y: {int(detection.spatialCoordinates.y)} mm", (x1 + 10, y1 + 65), cv2.FONT_HERSHEY_TRIPLEX, 0.5, 255)
            cv2.putText(frame, f"Z: {int(detection.spatialCoordinates.z)} mm", (x1 + 10, y1 + 80), cv2.FONT_HERSHEY_TRIPLEX, 0.5, 255)

            cv2.rectangle(frame, (x1, y1), (x2, y2), color, cv2.FONT_HERSHEY_SIMPLEX)

        cv2.putText(frame, "NN fps: {:.2f}".format(fps), (2, frame.shape[0] - 4), cv2.FONT_HERSHEY_TRIPLEX, 0.4, color)
        cv2.imshow("depth", depthFrameColor)
        cv2.imshow("rgb", frame)

        if cv2.waitKey(1) == ord('q'):
            break

Also available on GitHub

#include <chrono>

#include "utility.hpp"

// Includes common necessary includes for development using depthai library
#include "depthai/depthai.hpp"

static const std::vector<std::string> labelMap = {
    "person",        "bicycle",      "car",           "motorbike",     "aeroplane",   "bus",         "train",       "truck",        "boat",
    "traffic light", "fire hydrant", "stop sign",     "parking meter", "bench",       "bird",        "cat",         "dog",          "horse",
    "sheep",         "cow",          "elephant",      "bear",          "zebra",       "giraffe",     "backpack",    "umbrella",     "handbag",
    "tie",           "suitcase",     "frisbee",       "skis",          "snowboard",   "sports ball", "kite",        "baseball bat", "baseball glove",
    "skateboard",    "surfboard",    "tennis racket", "bottle",        "wine glass",  "cup",         "fork",        "knife",        "spoon",
    "bowl",          "banana",       "apple",         "sandwich",      "orange",      "broccoli",    "carrot",      "hot dog",      "pizza",
    "donut",         "cake",         "chair",         "sofa",          "pottedplant", "bed",         "diningtable", "toilet",       "tvmonitor",
    "laptop",        "mouse",        "remote",        "keyboard",      "cell phone",  "microwave",   "oven",        "toaster",      "sink",
    "refrigerator",  "book",         "clock",         "vase",          "scissors",    "teddy bear",  "hair drier",  "toothbrush"};

static std::atomic<bool> syncNN{true};

int main(int argc, char** argv) {
    using namespace std;
    using namespace std::chrono;
    std::string nnPath(BLOB_PATH);

    // If path to blob specified, use that
    if(argc > 1) {
        nnPath = std::string(argv[1]);
    }

    // Print which blob we are using
    printf("Using blob at path: %s\n", nnPath.c_str());

    // Create pipeline
    dai::Pipeline pipeline;

    // Define sources and outputs
    auto camRgb = pipeline.create<dai::node::ColorCamera>();
    auto spatialDetectionNetwork = pipeline.create<dai::node::YoloSpatialDetectionNetwork>();
    auto monoLeft = pipeline.create<dai::node::MonoCamera>();
    auto monoRight = pipeline.create<dai::node::MonoCamera>();
    auto stereo = pipeline.create<dai::node::StereoDepth>();

    auto xoutRgb = pipeline.create<dai::node::XLinkOut>();
    auto xoutNN = pipeline.create<dai::node::XLinkOut>();
    auto xoutDepth = pipeline.create<dai::node::XLinkOut>();
    auto nnNetworkOut = pipeline.create<dai::node::XLinkOut>();

    xoutRgb->setStreamName("rgb");
    xoutNN->setStreamName("detections");
    xoutDepth->setStreamName("depth");
    nnNetworkOut->setStreamName("nnNetwork");

    // Properties
    camRgb->setPreviewSize(416, 416);
    camRgb->setResolution(dai::ColorCameraProperties::SensorResolution::THE_1080_P);
    camRgb->setInterleaved(false);
    camRgb->setColorOrder(dai::ColorCameraProperties::ColorOrder::BGR);

    monoLeft->setResolution(dai::MonoCameraProperties::SensorResolution::THE_400_P);
    monoLeft->setCamera("left");
    monoRight->setResolution(dai::MonoCameraProperties::SensorResolution::THE_400_P);
    monoRight->setCamera("right");

    // setting node configs
    stereo->setDefaultProfilePreset(dai::node::StereoDepth::PresetMode::HIGH_DENSITY);
    // Align depth map to the perspective of RGB camera, on which inference is done
    stereo->setDepthAlign(dai::CameraBoardSocket::CAM_A);
    stereo->setOutputSize(monoLeft->getResolutionWidth(), monoLeft->getResolutionHeight());

    spatialDetectionNetwork->setBlobPath(nnPath);
    spatialDetectionNetwork->setConfidenceThreshold(0.5f);
    spatialDetectionNetwork->input.setBlocking(false);
    spatialDetectionNetwork->setBoundingBoxScaleFactor(0.5);
    spatialDetectionNetwork->setDepthLowerThreshold(100);
    spatialDetectionNetwork->setDepthUpperThreshold(5000);

    // yolo specific parameters
    spatialDetectionNetwork->setNumClasses(80);
    spatialDetectionNetwork->setCoordinateSize(4);
    spatialDetectionNetwork->setAnchors({10, 14, 23, 27, 37, 58, 81, 82, 135, 169, 344, 319});
    spatialDetectionNetwork->setAnchorMasks({{"side26", {1, 2, 3}}, {"side13", {3, 4, 5}}});
    spatialDetectionNetwork->setIouThreshold(0.5f);

    // Linking
    monoLeft->out.link(stereo->left);
    monoRight->out.link(stereo->right);

    camRgb->preview.link(spatialDetectionNetwork->input);
    if(syncNN) {
        spatialDetectionNetwork->passthrough.link(xoutRgb->input);
    } else {
        camRgb->preview.link(xoutRgb->input);
    }

    spatialDetectionNetwork->out.link(xoutNN->input);

    stereo->depth.link(spatialDetectionNetwork->inputDepth);
    spatialDetectionNetwork->passthroughDepth.link(xoutDepth->input);
    spatialDetectionNetwork->outNetwork.link(nnNetworkOut->input);

    // Connect to device and start pipeline
    dai::Device device(pipeline);

    // Output queues will be used to get the rgb frames and nn data from the outputs defined above
    auto previewQueue = device.getOutputQueue("rgb", 4, false);
    auto detectionNNQueue = device.getOutputQueue("detections", 4, false);
    auto depthQueue = device.getOutputQueue("depth", 4, false);
    auto networkQueue = device.getOutputQueue("nnNetwork", 4, false);

    auto startTime = steady_clock::now();
    int counter = 0;
    float fps = 0;
    auto color = cv::Scalar(255, 255, 255);
    bool printOutputLayersOnce = true;

    while(true) {
        auto imgFrame = previewQueue->get<dai::ImgFrame>();
        auto inDet = detectionNNQueue->get<dai::SpatialImgDetections>();
        auto depth = depthQueue->get<dai::ImgFrame>();
        auto inNN = networkQueue->get<dai::NNData>();

        if(printOutputLayersOnce && inNN) {
            std::cout << "Output layer names: ";
            for(const auto& ten : inNN->getAllLayerNames()) {
                std::cout << ten << ", ";
            }
            std::cout << std::endl;
            printOutputLayersOnce = false;
        }

        cv::Mat frame = imgFrame->getCvFrame();
        cv::Mat depthFrame = depth->getFrame();  // depthFrame values are in millimeters

        cv::Mat depthFrameColor;
        cv::normalize(depthFrame, depthFrameColor, 255, 0, cv::NORM_INF, CV_8UC1);
        cv::equalizeHist(depthFrameColor, depthFrameColor);
        cv::applyColorMap(depthFrameColor, depthFrameColor, cv::COLORMAP_HOT);

        counter++;
        auto currentTime = steady_clock::now();
        auto elapsed = duration_cast<duration<float>>(currentTime - startTime);
        if(elapsed > seconds(1)) {
            fps = counter / elapsed.count();
            counter = 0;
            startTime = currentTime;
        }

        auto detections = inDet->detections;

        for(const auto& detection : detections) {
            auto roiData = detection.boundingBoxMapping;
            auto roi = roiData.roi;
            roi = roi.denormalize(depthFrameColor.cols, depthFrameColor.rows);
            auto topLeft = roi.topLeft();
            auto bottomRight = roi.bottomRight();
            auto xmin = (int)topLeft.x;
            auto ymin = (int)topLeft.y;
            auto xmax = (int)bottomRight.x;
            auto ymax = (int)bottomRight.y;
            cv::rectangle(depthFrameColor, cv::Rect(cv::Point(xmin, ymin), cv::Point(xmax, ymax)), color, cv::FONT_HERSHEY_SIMPLEX);

            int x1 = detection.xmin * frame.cols;
            int y1 = detection.ymin * frame.rows;
            int x2 = detection.xmax * frame.cols;
            int y2 = detection.ymax * frame.rows;

            uint32_t labelIndex = detection.label;
            std::string labelStr = to_string(labelIndex);
            if(labelIndex < labelMap.size()) {
                labelStr = labelMap[labelIndex];
            }
            cv::putText(frame, labelStr, cv::Point(x1 + 10, y1 + 20), cv::FONT_HERSHEY_TRIPLEX, 0.5, 255);
            std::stringstream confStr;
            confStr << std::fixed << std::setprecision(2) << detection.confidence * 100;
            cv::putText(frame, confStr.str(), cv::Point(x1 + 10, y1 + 35), cv::FONT_HERSHEY_TRIPLEX, 0.5, 255);

            std::stringstream depthX;
            depthX << "X: " << (int)detection.spatialCoordinates.x << " mm";
            cv::putText(frame, depthX.str(), cv::Point(x1 + 10, y1 + 50), cv::FONT_HERSHEY_TRIPLEX, 0.5, 255);
            std::stringstream depthY;
            depthY << "Y: " << (int)detection.spatialCoordinates.y << " mm";
            cv::putText(frame, depthY.str(), cv::Point(x1 + 10, y1 + 65), cv::FONT_HERSHEY_TRIPLEX, 0.5, 255);
            std::stringstream depthZ;
            depthZ << "Z: " << (int)detection.spatialCoordinates.z << " mm";
            cv::putText(frame, depthZ.str(), cv::Point(x1 + 10, y1 + 80), cv::FONT_HERSHEY_TRIPLEX, 0.5, 255);

            cv::rectangle(frame, cv::Rect(cv::Point(x1, y1), cv::Point(x2, y2)), color, cv::FONT_HERSHEY_SIMPLEX);
        }

        std::stringstream fpsStr;
        fpsStr << std::fixed << std::setprecision(2) << fps;
        cv::putText(frame, fpsStr.str(), cv::Point(2, imgFrame->getHeight() - 4), cv::FONT_HERSHEY_TRIPLEX, 0.4, color);

        cv::imshow("depth", depthFrameColor);
        cv::imshow("rgb", frame);

        int key = cv::waitKey(1);
        if(key == 'q' || key == 'Q') {
            return 0;
        }
    }
    return 0;
}

Got questions?

Head over to Discussion Forum for technical support or any other questions you might have.