Mastering OpenCV 4 with Python

MobileNet-SSD for object detection

We are going to use a MobileNet architecture combined with an SSD framework. MobileNets can be seen as efficient convolutional neural networks for mobile vision applications.

The MobileNet-SSD was trained on the COCO dataset and fine-tuned on PASCAL VOC reaching 72.27% mAP (see the table summarizing mAP for object detection algorithms to put this metric in context). When fine-tuned on PASCAL VOC, 20 object classes can be detected as follows:

• Person: Person
• Animal: Bird, cat, cow, dog, horse, and sheep
• Vehicle: Aeroplane, bicycle, boat, bus, car, motorbike, and train
• Indoor: Bottle, chair, dining table, potted plant, sofa, and TV/monitor

In the object_detection_opencv_mobilenet_caffe.py script, we perform ...