Video demo

[MuhammadAsadJaved]

Hi, Can you please adjust the code for running demo on video inputs?

Thanks

[CalayZhou]

Hi, Can you please adjust the code for running demo on video inputs?

Thanks

hello, you can change the function test_MBNet() in MBNetModel.py a little to show the video demo:

def test_MBNet(self,opt, data_path,val_data, weight_path):
    self.model_all.load_weights(weight_path, by_name=True)
    print ('load weights from {}'.format(weight_path))
    val_data.sort(reverse =False)
    for f in range(len(val_data)):
        img_name = os.path.join(data_path,val_data[f])
        if not img_name.lower().endswith(('.jpg', '.png')):
            continue
        print(img_name)
        img = cv2.imread(img_name)

        img_name_lwir = os.path.join(data_path[:-7]+'lwir', val_data[f][:-11]+'lwir.png')
        print(img_name_lwir)
        img_lwir = cv2.imread(img_name_lwir)

        start_time = time.time()
        x_in = bbox_process.format_img(img, opt)
        x_in_lwir = bbox_process.format_img(img_lwir, opt)
        Y = self.model_all.predict([x_in,x_in_lwir])
        proposals = bbox_process.pred_pp_1st(self.anchors, Y[0], Y[1], opt)
        bbx, scores = bbox_process.pred_det(proposals, Y[2], Y[3], opt, step=2)
        print ('Test time: %.4f s' % (time.time() - start_time))

        for ind in range(len(bbx)):
            if scores[ind][0] < 0.5:
                continue
            (x1, y1, x2, y2) = bbx[ind, :]
            cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)
            # cv2.putText(img, str(scores[ind][0]), (x1, y1), cv2.FONT_HERSHEY_COMPLEX, 0.3, (0, 0, 255), 1)
            cv2.rectangle(img_lwir, (x1, y1), (x2, y2), (0, 255, 0), 2)
        # cv2.putText(img_lwir, str(scores[ind][0]), (x1, y1), cv2.FONT_HERSHEY_COMPLEX, 0.3, (0, 0, 255), 1)
        img_concat = np.concatenate([img, img_lwir], axis=1)

        cv2.imshow("img_concat",img_concat)
        cv2.waitKey(10)

then python test.py and the video demo will be shown.

[MuhammadAsadJaved]

@CalayZhou Hi dear, I think your suggested program is still for the images as your input is image sequences. I have changed the script for the video demo and also added demo_video.py accordingly. I have sent you a pull request, please review it and approve if it's useful.

[CalayZhou]

@MuhammadAsadJaved I think demo_video.py is reasonable and helpful, i have merged it with the MBNet-master, Thanks for your contribution.