My understanding is that Faster RCNN is an architecture for performing object detection. It finds objects in an image and classifies them. My understanding is also that VGG-16, RESNET-50, etc... also find objects in images and classify them. Are VGG-16, RESNET-50, etc... object detection architectures in the same way that Faster RCNN is? I often see VGG-16, RESNET-50, etc... as the "backbone" for Faster RCNN and am seriously confused by the literature. Thanks in advance!
Image Classification Models are commonly referred as a combination of feature extraction and classification sub-modules. Where the total model excluding last layer is called feature extractor, and the last layer is called classifier. Popular Image Classification Models are: Resnet, Xception, VGG, Inception, Densenet and Mobilenet.
Object Detection Models are more combination of different sub-models rather than single end to end connected models, as you mentioned it is more like an architecture. Object detection model contains a feature extraction model, region proposal network, classification and regression models. Here the above mentioned classification models (Resnet50, VGG, etc) excluding all dense layers are used as a feature extractors.
Some of the popular object detection models are: Region-based Convolutional Neural Networks(R-CNN), Spatial Pyramid Pooling(SPP-net), Fast R-CNN, Faster R-CNN, Regression-based object detectors, YOLO(You only Look Once),Single Shot Detector(SSD).
Short answer - No
Let me put it this way - let's say you have some image dataset. You need to perform any deep learning task on it(classification, object detection, segmentation etc).
For this purpose these are the steps you'll be performing in general :
- Define labels for each input
- Define how will you extract feature
- Define how will you use the features and labels to perform supervised learning
- Define loss function and optimiser
- Define training and testing pipeline
VGG16, ResNet16 etc are architectures. They perform the 2nd step - feature extraction.
Faster-RCNN means all of 1-5 steps defined specifically for object detection task.