bounding box of videostream. visualized in streamlit

bbox-video-streamlit.py

#%%
import time
import numpy as np
import streamlit as st
import pandas as pd
import cv2

video_sources = {
    'Rindabotn' : "http://217.17.213.66:8081/mjpg/video.mjpg?overview=0&camera=1&videoframeskipmode=empty&Axis-Orig-Sw=true&resolution=1280x720",
    'Hodlekve' : "http://217.17.213.66:8080/mjpg/video.mjpg?overview=0&camera=1&videoframeskipmode=empty&Axis-Orig-Sw=true&resolution=1280x720",
    'Times Square ground' : 'https://videos3.earthcam.com/fecnetwork/9974.flv/chunklist_w1421640637.m3u8',
    'Times Square' : 'https://videos3.earthcam.com/fecnetwork/hdtimes10.flv/chunklist_w1581551174.m3u8',
    'Road' : 'https://videos3.earthcam.com/fecnetwork/15559.flv/chunklist_w1295434004.m3u8'
}

#%% Run the YOLO model to detect objects.
# Borrowed functions from https://github.com/streamlit/demo-self-driving/blob/master/app.py
def yolo_v3(image, confidence_threshold, overlap_threshold):
    # Load the network. Because this is cached it will only happen once.
    @st.cache(allow_output_mutation=True)
    def load_network(config_path, weights_path):
        net = cv2.dnn.readNetFromDarknet(config_path, weights_path)
        output_layer_names = net.getLayerNames()
        output_layer_names = [output_layer_names[i[0] - 1] for i in net.getUnconnectedOutLayers()]
        return net, output_layer_names
    net, output_layer_names = load_network("yolov3.cfg", "yolov3.weights")

    # Run the YOLO neural net.
    blob = cv2.dnn.blobFromImage(image, 1 / 255.0, (416, 416), swapRB=True, crop=False)
    net.setInput(blob)
    layer_outputs = net.forward(output_layer_names)

    # Supress detections in case of too low confidence or too much overlap.
    boxes, confidences, class_IDs = [], [], []
    H, W = image.shape[:2]
    for output in layer_outputs:
        for detection in output:
            scores = detection[5:]
            classID = np.argmax(scores)
            confidence = scores[classID]
            if confidence > confidence_threshold:
                box = detection[0:4] * np.array([W, H, W, H])
                centerX, centerY, width, height = box.astype("int")
                x, y = int(centerX - (width / 2)), int(centerY - (height / 2))
                boxes.append([x, y, int(width), int(height)])
                confidences.append(float(confidence))
                class_IDs.append(classID)
    indices = cv2.dnn.NMSBoxes(boxes, confidences, confidence_threshold, overlap_threshold)

    # Map from YOLO labels to Udacity labels.
    UDACITY_LABELS = {
        0: 'pedestrian',
        1: 'biker',
        2: 'car',
        3: 'biker',
        5: 'truck',
        7: 'truck',
        9: 'trafficLight'
    }
    xmin, xmax, ymin, ymax, labels = [], [], [], [], []
    if len(indices) > 0:
        # loop over the indexes we are keeping
        for i in indices.flatten():
            label = class_IDs[i]#UDACITY_LABELS.get(class_IDs[i], None)
            if label is None:
                continue

            # extract the bounding box coordinates
            x, y, w, h = boxes[i][0], boxes[i][1], boxes[i][2], boxes[i][3]

            xmin.append(x)
            ymin.append(y)
            xmax.append(x+w)
            ymax.append(y+h)
            labels.append(label)

    boxes = pd.DataFrame({"xmin": xmin, "ymin": ymin, "xmax": xmax, "ymax": ymax, "labels": labels})
    return boxes[["xmin", "ymin", "xmax", "ymax", "labels"]]
def create_image_with_boxes(image, boxes, header, description):
    # Superpose the semi-transparent object detection boxes.    # Colors for the boxes
    LABEL_COLORS = {
        "car": [255, 0, 0],
        "pedestrian": [0, 255, 0],
        "truck": [0, 0, 255],
        "trafficLight": [255, 255, 0],
        "biker": [255, 0, 255],
    }
    image_with_boxes = image.astype(np.float64)
    for _, (xmin, ymin, xmax, ymax, label) in boxes.iterrows():
        image_with_boxes[int(ymin):int(ymax),int(xmin):int(xmax),:] += [label*10,255-10*label,label] #LABEL_COLORS[label]
        image_with_boxes[int(ymin):int(ymax),int(xmin):int(xmax),:] /= 2
    return image_with_boxes.astype(np.uint8)

#%%
import cv2
from threading import Thread


class ThreadedCamera(object):
    def __init__(self, source = 0):

        self.capture = cv2.VideoCapture(source)

        self.thread = Thread(target = self.update, args = ())
        self.thread.daemon = True
        self.thread.start()

        self.status = False
        self.frame  = None

    def update(self):
        while True:
            if self.capture.isOpened():
                (self.status, self.frame) = self.capture.read()

    def grab_frame(self):
        if self.status:
            return self.frame
        return None  
#%% DISPLAY

st.sidebar.title("Parameters")
source = st.sidebar.radio("Velg kamera", list(video_sources.keys()))
st.markdown(f"# {source}")
url = video_sources[source]
streamer = ThreadedCamera(url)
cap = cv2.VideoCapture(url)

confidence_threshold = st.sidebar.slider("Confidence threshold", 0.0, 1.0, 0.5, 0.01)
overlap_threshold = st.sidebar.slider("Overlap threshold", 0.0, 1.0, 0.3, 0.01)
sleeptime = st.sidebar.slider("Sleep", 0.0, 10.0, 1.0, 0.1)
stream = st.empty()

while True:
    #result, frame = cap.read()
    frame = streamer.grab_frame()
    boxes = yolo_v3(frame, confidence_threshold=confidence_threshold, overlap_threshold=overlap_threshold)
    frame = create_image_with_boxes(frame,boxes, header="",description="")
    stream.image(frame, use_column_width=True)
    time.sleep(sleeptime)