def video_demo(args): cap = cv2.VideoCapture('data/demo_video.mp4') while (cap.isOpened()): ret, frame_bgr = cap.read() if not ret: break with Tick('interference'): detections = predict(frame_bgr) overlap = plot(detections, frame_bgr) cv2.imshow('overlap', overlap) if 27 == cv2.waitKey(1): break
def image_callback(self,ros_data): if self.input_compressed: np_arr = np.fromstring(ros_data.data, np.uint8) img_input = cv2.imdecode(np_arr,1) else: img_input = bridge.imgmsg_to_cv2(ros_data) with Tick('interference'): detections = predict(img_input) overlap = plot(detections,img_input) detections = overlap cv2.imshow('overlap',overlap) cv2.waitKey(1)
def images_demo(args): for fn in glob.glob('data/COCO/*'): frame_bgr = cv2.imread(fn) with Tick('interference'): detections = predict(frame_bgr) rcnn_overlap = plot(detections, frame_bgr) fig = plt.figure(figsize=(12, 4), dpi=100, facecolor='w', edgecolor='k') sub_plot(fig, 1, 2, 1, 'image', cv2.cvtColor(frame_bgr, cv2.COLOR_BGR2RGB)) sub_plot(fig, 1, 2, 2, 'overlap', cv2.cvtColor(rcnn_overlap, cv2.COLOR_BGR2RGB)) plt.show(block=False) plt.show()
import os os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' os.environ['CUDA_VISIBLE_DEVICES'] = '0' import time import cv2 import json import glob import numpy as np import matplotlib.pyplot as plt import random import keras import tensorflow as tf from tensorflow.python.keras import backend as K from model_wrapper.utils import sub_plot,Tick,voc from mask_rcnn.warpper import predict,plot cap = cv2.VideoCapture('data/demo_video.mp4') while(cap.isOpened()): ret, frame_bgr = cap.read() if not ret: break with Tick('interference'): detections = predict(frame_bgr) overlap = plot(detections,frame_bgr) cv2.imshow('overlap',overlap) if 27 == cv2.waitKey(1): break